Estimating the Prevalence of Injection Drug Use in Five Cities of Georgia

"Programme of Assistance for the Prevention of Drug Abuse and Drug Trafficking in the Southern Caucasus (SCAD Programme). This Programme is funded by the European Union"

This document was prepared by Bemoni Public Union

with support from SCAD Programme

This publication was made possible through the support provided by the "Programme of Assistance for the Prevention of Drug Abuse and Drug Trafficking in the Southern Caucasus (SCAD-5 Programme). This Programme is funded by the European Union" The opinions expressed herein are those of the authors' and do not necessarily reflect the views of EU 1

REPORT PREPARED BY

Tamar Sirbiladze, M.D., Expert Team Leader, Chairperson of the Bemoni Public Union

Lela Tavzarashvili, M.D., Project Coordinator,

Tomas Zabransky, M.D., Ph.D., International Epidemiology Expert, SCAD Programme

Lela Sturua, Ph.D., Epidemiology Expert, SCAD Programme

ACKNOWLEDGEMENTS

The development of this report became possible due to financial support of the European Union. We wish to express our deep appreciation to the Global Fund Project to Fight AIDS, Tuberculosis and Malaria and Curatio International Foundation for providing us the opportunity to collaborate in this venture.

Writing this report was a huge task, and its completion was made possible by the help of SCAD epidemiology experts. First, recognition should go to Prof. Tomas Zabransky, International Epidemiology Expert of SCAD Programme for his efforts, recommendations, instructive and valuable comments.

The authors thank the project staff. It was their hard work, determination and dedication, which resulted in timely completion of this study.

Special thanks are extended to experts Mrs. Khatuna Todadze, Deputy Director of the Research Institute on Addiction, Mr. Akaki Gamkrelidze, Director of the Drug Prevention Center within the National Center for Disease Control and Public Health (NCDC), Mrs. Ketevan Shermadini, epidemiologist of the Infectious Diseases, AIDS and Clinical Immunology Research Center, and Mr. Tamaz Zakarashvili, representative from the Ministry of Internal Affairs. Without their contributions, the report would not have acquired its present shape.

We also wish to acknowledge the inputs of all our implementing NGO partners in regions for their support during field implementation. The database and statistic processing specialists, who created the prevalence study database, were extremely flexible and professional.

Many individuals have provided direct and/or indirect support to bring this report to successful completion. We thank all of them, even if they are not mentioned individually by name.

The authors are especially appreciative of the injection drug users who were willing to dedicate their time to participating in the study.

ACRONYMS

AIDS – Acquired Immune Deficiency Syndrome AIDS Center – Infectious Diseases, AIDS & Clinical Immunology Research Center BPU – Bemoni Public Union BSS – Behavioral Surveillance Survey CI - Confidence interval CIF – Curatio International Foundation ELISA – Enzyme Linked Immunosorbent Assay EMCDDA – European Monitoring Center for Drugs and Drug Addiction EU – European Union FGD - Focus Group Discussion GEL – Georgian Lari (exchange rate of 1.66GEL = 1USD at the time of this report) GFATM – Global Fund to Fight AIDS, Tuberculosis and Malaria GoG – Government of Georgia GRIA– Georgian Research Institute on Addiction HIV – Human Immunodeficiency Virus ICD-10 – International Statistical Classification of Diseases and Related Health Problems, 10th revision IDUs – Injecting Drug Users IEC – Information, Education, Communication MARP - Most at risk population MoIA - Ministry of Internal Affairs of Georgia MoLHSA – Ministry of Labor, Health and Social Affairs of Georgia NCDC - National Center for Disease Control and Public Health NGO – Non-Government Organization PCR - Polymerase Chain Reaction RDS – Respondent Driven Sampling RDSCM - Respondent Driven Sampling Coupon Management RDSAT - Respondent Driven Sampling Analysis Tool RPR – Rapid Plasma Reagent SCAD – South Caucasus Anti-drug Programme SHIP – STI/HIV Prevention SPSS – Statistical Package for the Social Sciences STI – Sexually Transmitted Infections TPHA – Treponema pallidum Hemaggultination Assay UNAIDS – The Joint United Nations Programme on AIDS UNDP – United Nations Development Program UNODC - United Nations Office on Drugs and Crime USAID – United States Agency for International Development WB – Western Blot Test WHO – World Health Organization VCT – Voluntary Counseling and Testing

DEFINITIONS

Population: the entire group of individuals or items of interest in the study.

Target population: the population from which representative information is desired and to which inferences will be made.

The prevalence of a certain social attribute is defined as the proportion of people possessing that attribute. It is often expressed as a percentage, or sometimes as “per thousand” or even “per million” of the total population. The actual number of individuals is sometimes used instead of the prevalence, however without information on the baseline population, this number may be meaningless.

Prevalence is a measure of how many drug users there are in a community or country and how they are distributed across the population e.g. by age, gender, geographical location or type of drug use.

The term “Lifetime Prevalence” refers to: the proportion of the population who have used a particular drug at least once, whereas “Current Prevalence” refers to those who have used a particular drug in a specific period of time such as the last month/week.

According to EMCDDA1 definition, Problem drug use (PDU) is defined as “Injecting drug use or long duration/regular use of opioids, cocaine and/or amphetamines”. Amphetamines include both amphetamine and methamphetamine, but not ecstasy. Opioids include any legal or illegal use of any opioids (e.g. methadone, buprenorphine, slow release morphine).

Injecting drug use (IDU) is defined as “Injecting for non-medical purposes”.

Indicators are data which give pointers or act as tools in the estimation of prevalence e.g. data collected routinely by government agencies such as arrest data, drug treatment data and mortality data. The data reflects only those who have come into contact with services and not all users of illicit drugs.

Estimation methods are the range of methods which can be used to estimate the prevalence of illicit drug use.

Routine data sources - statistics that are collected routinely i.e. in the course of duty.

Non-routine data sources - statistics that are not routinely collected but are “once-offs” such as the results of studies of drug use in the general population or in a specific group. These can be gathered for research or planning purposes.

BSS is the ongoing systematic collection, analysis and interpretation of HIV/AIDS data and the dissemination of information to those who need to know so that actions may be taken.

Sampling is the process of selecting a portion of a population in order to make inferences about the larger population from which the sample was drawn. Sampling is of crucial importance in measuring trends over time. Sampling strategies should therefore be systemic and replicable over time.

1 European Monitoring center on Drugs and Drug Addiction 4

Table of Contents EXECUTIVE SUMMARY ...... 8 Study Design and Methods ...... 8 Key Findings ...... 9 Extrapolation from Local to National Prevalence Estimates ...... 9 Conclusion ...... 10 INTRODUCTION ...... 12 Overview of Drug Situation in Georgia ...... 12 How many IDUs are there in the country? ...... 13 Brief overview of size estimation methods ...... 15 Respondent-driven Sampling (RDS) as a Strategy to Reach Hard-to-reach Populations ...... 17 Complementary Projects ...... 18 METHODOLOGY ...... 19 Objective of the Study ...... 19 Defining the Target Population ...... 19 Stages of IDU prevalence estimation method to be applied for Georgia ...... 19 Limitation of the study ...... 21 Study Design ...... 22 Pre-study activities ...... 23 Geographical Scope ...... 23 Study sites and staff ...... 24 Ethical Considerations ...... 24 Steps in Data Collection ...... 25 RDS ...... 25 Benchmark Data Collection ...... 33 Data Entry and Analysis ...... 35 Quality Control ...... 36 FINDINGS ...... 37 1. RDS Survey ...... 37 2. Benchmark Data ...... 43 3. Calculation of the Size of IDU Population in 5 selected cities ...... 48 4. Estimation of the prevalence of injection drug use ...... 50 EXTRAPOLATION FROM LOCAL TO NATIONAL PREVALENCE ESTIMATES ...... 52 Extrapolation Method: The multivariate indicator for injection drug use ...... 52 Application ...... 53 Results of the national prevalence estimation ...... 56 DISCUSSION ...... 60 CONCLUSION AND RECOMMENDATIONS ...... 63 5

REFERENCES ...... 65 APPENDIXES ...... 66 Appendix 1. Socio-demographic characteristics of non-respondents in research studies using nomination techniques ...... 66 Appendix 2. Key characteristics of IDU sample in five locations across the country ...... 67 Appendix 3. Nomination method/questionnaire ...... 77

List of Figures Figure 1. Estimated Prevalence of IDUs in Urban Areas of Georgia ...... 11 Figure 2. Study Design ...... 22 Figure 3. Map of Georgia with Study Sites ...... 24 Figure 4. RDS Recruitment Methods ...... 26 Figure 5. Sample of Recruitment Pattern in the Study (Tbilisi) ...... 30 Figure 6. Distribution of Respondents by Age Groups ...... 37 Figure 7. Distribution of Respondents by Educational Level ...... 38 Figure 8. Distribution of Respondents by Marital Status ...... 38 Figure 9. Age When First Used Any Drug...... 39 Figure 10. Age When First Injected Any Drug ...... 39 Figure 11. Injected Drugs Last Week ...... 41 Figure 12. Detoxification Treatment ...... 42 Figure 13. Other Services ...... 42

List of Tables Table 1. Basic multiplier-benchmark calculation ...... 20 Table 2. Socio-demographic characteristics of seeds vs. total sample in Tbilisi ...... 27 Table 3. Socio-demographic characteristics of seeds vs. total sample in Gori ...... 27 Table 4. Socio-demographic characteristics of seeds vs. total sample in Telavi ...... 28 Table 5. Socio-demographic characteristics of seeds vs. total sample in Zugdidi ...... 28 Table 6. Socio-demographic characteristics of seeds vs. total sample in Batumi ...... 29 Table 7. Ineligibility, refusals and drop-outs ...... 31 Table 8. Number of acquaintances nominated by IDUs ...... 32 Table 9. Mean duration of injecting drug use ...... 40 Table 10. Mean number of injecting group members ...... 40 Table 11. Prevalence of HIV and STIs ...... 43 Table 12. Detoxification treatment benchmark data ...... 43 Table 13. Methadone substitution treatment benchmark data ...... 44 Table 14. Methadone substitution waiting list benchmark data ...... 45 Table 15. Needle exchange benchmark data ...... 45 Table 16. VCT services benchmark data ...... 45 Table 17. HIV testing benchmark data ...... 46 Table 18. Benchmark data on IDUs came into contact with the police ...... 47 Table 19. Estimates of the number of IDUs in Tbilisi ...... 48 Table 20. Estimates of the number of IDUs in Gori ...... 48 Table 21. Estimates of the number of IDUs in Telavi ...... 49 Table 22. Estimates of the number of IDUs in Zugdidi ...... 49 Table 23. Estimates of the number of IDUs in Batumi ...... 49 Table 24. Estimated Prevalence Rates in Tbilisi ...... 50 Table 25. Estimated Prevalence Rates in Gori ...... 50 Table 26. Estimated Prevalence Rates in Telavi ...... 51 Table 27. Estimated Prevalence Rates in Zugdidi ...... 51

Table 28. Estimated Prevalence Rates in Batumi ...... 51 Table 29. Estimation by Population Density ...... 56 Table 30. Estimation by Prevalence Rate Coefficient ...... 57 Table 31. Characteristics of non-respondents in IDU prevalence study ...... 66 Table 32. Key Characteristics of IDU sample in Tbilisi ...... 67 Table 33. Key Characteristics of IDU sample in Gori ...... 69 Table 34. Key Characteristics of IDU sample in Telavi ...... 71 Table 35. Key Characteristics of IDU sample in Zugdidi ...... 73 Table 36. Key Characteristics of IDU sample in Batumi ...... 75

EXECUTIVE SUMMARY

Drug abuse and related health, social and economic consequences are a critical problem facing Georgia today. There are presently no valid data on the prevalence and incidence of drug abuse among the general population of Georgia. It is difficult to overestimate the importance of obtaining accurate information on the prevalence of illicit drug use.

A variety of methods are available for estimating the prevalence of heavier or more problematic patterns of illegal drug use. Of all the methods of indirect estimation the multiplier-benchmark approach is probably the easiest to implement and probably the one with the longest history of use in the field of drug epidemiology.

This is the first time the multiplier/benchmark method has been applied to estimate an Injection Drug Use (IDU) population in Georgia. For the purpose of this study, we regarded any person who has used any psychoactive drug through injections in a non-medical context.

Study Design and Methods The aim of the present study is to estimate the prevalence of Injection Drug Use (IDU) in Tbilisi (the capital) and 4 main cities (Batumi, Telavi, Gori and Zugdidi) of Georgia and provide IDU prevalence estimate throughout the Country.

There are five stages of prevalence estimation method that had been used in this study. Stage 1: Data collection of IDUs (gaining the benchmark data - B) - all available data on injection drug use in Georgia were reviewed. Data of IDUs are recorded under the current system for the year 2007 (details see below in chapter “Benchmark Data Collection”). Stage 2: Estimation of the value of multiplier (M) - the proportion of the target population in the benchmarks is obtained from research studies using nomination techniques (study using the Respondent Driven Sampling (RDS) methodology based on appropriate eligibility criteria and accurate sample size calculations was conducted). The survey collected the data among IDUs using nomination method/questionnaire developed by SCAD epidemiology experts. Stage 3: The derivation of multiplier - this stage involves two steps: a) Estimation of the percentage (P) of IDUs recorded from Stage 2. Separate estimates for different benchmarks were made in each city. b) Multiplier (M) is estimated for each benchmark by the inverse of percentages (Pisani, 2002). The formula M = 100/P Stage 4: Estimate the number of drug injectors - numbers of IDUs estimates for each benchmark are obtained by multiplying the recorded number of IDUs (collected from the available data source) by an appropriate multiplier (The formula E = BxM). Stage 5: Calculation of a prevalence of drug injection for each city - it was based on data on population distribution (State Department of Statistics of the Ministry of Economic Development of Georgia). Census data gave the population for urban areas. The population between 18 and 65 was used as the denominator for the prevalence based estimate. The appropriate estimates of injecting drug use were then applied to that adult population. An upper and lower limit is provided by statistical means. Additionally, the first attempt to derive the national estimate for the percentage of injection drug users in Georgia using the Multiple Indicator Method (MIM) had been carried out.

Since 2008, program entitled “Establishment of Evidence-based Basis for HIV/AIDS National Program by Strengthening Surveillance System” is being implementing within the framework of the Global Fund Project to Fight AIDS, Tuberculosis and Malaria (GFATM) by Curatio international

Foundation (CIF) in cooperation with AIDS and Clinical Immunology Research Center, National Center for Disease Control and Public Health (NCDC) and local NGOs Bemoni and Tanadgoma. Under this project Bemoni proposed to conduct Behavioral Surveillance Surveys (BSSs) with a Biomarker Component among injecting drug users (IDUs) in five main cities of Georgia: Tbilisi, Gori, Telavi, Zugdidi, and Batumi. It was decided to use this opportunity and incorporate the nomination study for estimating the size of the injecting drug user (IDU) population into the above mentioned BSSs.

Key Findings Calculation of the Size of IDU Population. Multipliers were derived from the RDS survey of 1127 IDUs recruited from across 5 cities. Participants’ responses to the questionnaire were used to produce a final series of IDU size estimates, including 95% confidence intervals.

The population size estimate for IDUs was the mean of six multiplier estimations in Tbilisi, 5 in Batumi, 3 in Gori and Zugdidi and 2 in Telavi. This study suggests using the statistical lower and upper limits (at 95% confidence interval) to reflect the minimum and maximum ranges.

Estimation of the prevalence of injection drug use. Prevalence estimates for the injection drug use were produced for 5 cities of Georgia. Census data gave the population between 18 and 64 for urban areas across the country. The statistical lower and upper limits (at 95% confidence interval) were used to reflect the minimum and maximum ranges.

Calculation of the IDU prevalence estimation in the surveyed cities revealed these figures (mean estimates): Tbilisi – 4,03 (3,98-4,09); Gori – 3,61 (3,47-3,75); Telavi – 1,30 (1,19-1,42); Zugdidi: 4,63 (4,37-4,76); Batumi – 7,97 (7,79-8,15).

Extrapolation from Local to National Prevalence Estimates Local estimates using multiplier-benchmark methods give important information on extent of drug problem. However, they are employed in studies of drug use on a smaller, geographically local scale. Nonetheless, there is still very often a need for overall national estimates to be made, and one way of doing that is to extrapolate from local prevalence studies to an overall picture. The extrapolation methods are based on statistical regression techniques.

The Multivariative Indicator Method (MIM) had been used to derive national prevalence estimates. The aim of this method is to estimate the number of injection drug users in the population by combining information on prevalence that is available only in a few areas (the calibration population, or anchor points) and indicators or predictors of drug use that are available in all areas.

Two separate national estimations were produced: at first, national IDU prevalence was calculated using demographic indicator such as population density and the second method used the drug injection prevalence rate coefficient for each city.

National prevalence estimates for the injection drug use were produced for 65 cities of Georgia. Calculation of the IDU prevalence estimation nationwide revealed these figures: estimation method

N 1, using demographic indicator (population density) – 1,46% (estimated number of IDUs equals 39 152) estimation method N 2, using prevalence rate coefficients - 1,53% (Number of IDUs – 41 062).

Conclusion In this report we have provided the first ever estimates of the prevalence of injection drug use within Georgia. We have shown that injection drug use is occurring in all selected cities of Georgia and that, on average, from 1, 30% to 7, 97% of the population aged between 18 and 64 has used these drugs within the year 2007. Since this is the first time that an estimate of this kind has been produced there is no previous figure with which to make comparisons. The current study has demonstrated that it is feasible to apply the multiplier-benchmark method to the task of estimating the size of IDU population in Georgia.

The recording of information on problem drug use should be improved. The treatment monitoring system should not only provide figures of drug users seeking treatment categorized by main substance groups, but should also be able to avoid double counting.

Establishment of the Unique Identifier Code (UIC) system of anonymous client registration and tracking service is required. Therefore the actual time and effort spent collecting data will be reduced and this would further minimize the costs of a prevalence estimation exercise in the future. Thus when sufficient data have been collated, methods such as the truncated Poisson method or the capture-recapture method can be used to provide prevalence estimations.

The best results were found for police multiplier and treatment coverage (both detoxification and substitution) multiplier methods. They offer rather stable estimates. The police multiplier method is based on the number of individuals registered as drug offenders. The treatment coverage multiplier is based on the number of individuals treated for addiction problems that had in contact with treatment services in a given year period. Despite the perception that the estimate derived from HIV testing data within a multiplier method may be an underestimate, this method appears to be the most suitable for estimating the size of injecting populations in Georgia, since this indicator is available across the country.

The multiple indicator method to derive national prevalence estimates is cost-effective, as it does not require new data collection, unless separate studies are needed to estimate new anchor points. Local estimation methods should be used and further developed to produce regional anchor points for the multivariative indicator method. The current study suggests that a more differentiated response to the problem of drug abuse may be possible, although more work is required to provide more detailed breakdowns in terms of drug-related and demographic indicators.

Finally, this research has shown that it is possible to provide estimates of the prevalence of problematic drug use at both a national and local level within Georgia. It will be important to build upon this work so that over time we have a much clearer picture of the extent to which the drug problem in Georgia is changing. We also have to recognize that the problem of illegal drugs within the country can change rapidly. This indicates the importance of developing accurate on-going monitoring systems to identify rapid changes in the behavior of drug users within Georgia. Similar studies should be conducted on the regular basis. Since, the technical expertise for conducting such exercises is limited in the country at this moment; there is a need to develop pool of experts at the national level.

Figure 1. Estimated Prevalence of IDUs in Urban Areas of Georgia

Tsageri Oni Kazbegi

Kareli Kaspi

Sagarejjo

Very High High Medium Low Very Low

INTRODUCTION Overview of Drug Situation in Georgia Georgia’s population is estimated to be approximately 4.4 million in a geographical area of 70,000- sq. km., bounded by the Black Sea, Russia, Azerbaijan, Armenia and Turkey. Drug abuse and related health, social and economic consequences are a critical problem facing Georgia today. Drug addiction has escalated in Georgia since 1990. On the one hand, the collapse of the Soviet Union was followed by a breakdown of the anti-drug system specific for the totalitarian state that was mainly based on prohibitive measures. On the other hand, the social, political and economic events unfolding in the country gave rise to a series of incentives for intensive abuse of drugs. In particular, uncontrolled territories, unprotected frontiers, a sharp deterioration in the criminal situation, and corruption all widened access to drugs. At the same time, the social-economic collapse, a crisis of value in the society, social pessimism, and unemployment promoted increased drug abuse. Although recent years have witnessed economic development and reduction of crime, illicit sale and abuse of drugs are still on the increase.

The situation is worsened by the geographic location of Georgia, turning the country into one of the important routes for transiting drugs from Asia to Europe. In this respect, Georgia have appeared as an immediate link between the routes through which drugs flow from Afghanistan and Central Asia to Europe. Part of the drugs remains in Georgia, facilitating their increased abuse in the country. Recent years have seen a sharp rise in the smuggling of drugs from Europe, particularly in respect of Subutex® (buprenorphine) the abuse of which in Georgia reached alarming dimensions in 2004-2005. This medical product, used for the purpose of substitution therapy by means of sublingual administration, is basically used through injections in Georgia.2

According to the information of the Research Institute on Addiction, the number of officially registered drug abusers and drug addicts increased 9-fold from 1990 to 2004. Cannabis (marijuana, pot) is on top of the list of non-injection drugs in the country, which is suggested by the data contained in the republican database till 2005, and youth survey results, conducted in 2005 by Georgian Research Institute on Addiction.3 According to survey results, the most frequently consumed drug, both for experimental and recreational purposes, is marijuana. 52.6% of male respondents took marijuana at least once or twice in their lifetime.

In recent years, the abuse of the marijuana has covered the entire territory of Georgia. Georgian climate favours the growth of Cannabis, which is one of the factors conducive to the rise in consumption. Locally manufactured marijuana is easy to procure, and it is cheap. Besides, recent years have seen indirect promotion of marijuana by certain representatives of media and the show business, which stimulates young people’s interest to marijuana, creates positive disposition and blunts caution.4 Injection drugs are more available in large cities and locally cultivated pot is particularly widely spread in villages.

The abuse of Ecstasy by young people also draws attention, particularly in large cities. The high level of sniffing various volatile substances of drug effect has reached alarming scales among children and adolescents, particularly children devoid of parental care. According to the Youth Survey, Ecstasy (4.5%) ranks first after marijuana, followed by heroin (3.9%), Tramadol (2.6%) and inhalants (2.6%).5

2 Drug Situatuion in Georgia, Annual Report 2005, SCAD 3 Drug Situatuion in Georgia, Annual Report 2004, SCAD 4 Drug Situatrion in Georgia, Annual Report 2003, SCAD 5 Drug Situatuion in Georgia, Annual Report 2005, SCAD 12

The use of cocaine and amphetamine is insignificant, as they are not actually available on the black market. Ephedrine and pervitine (methamphetamine), which are usually prepared through chemical refinement of medicines used against respiratory disorders and are available from drugstores without any prescription, have also appeared in the black market. The number of females constituted 1% of the overall number of registered drug users. In terms of age, most illicit drug users are 21 to 35 (Gamkrelidze et al., 2005). 6

The gamut of drug abusers in terms of age, social and geographic belonging has also widened. On the one hand, there is an upward trend in the rejuvenation of the age of the first abuse of drugs, including injection ones. On the other hand, the age range of drug abusers has widened from children to the elderly. Drugs are abused in almost all social classes.7

Although on account of cultural and traditional specifics, the abuse of drugs by women is not as intensive as it is in some other countries (e.g. Russia, Ukraine, some European countries, etc), experts point out that the number of drug abusing women and girls has significantly increased in recent years.

Drug usage and addiction is the primary driver of the HIV/AIDS epidemic in Georgia. Low awareness of HIV/AIDS, high levels of unsafe injecting coupled with unsafe sexual behaviour as well as high transmission rates of other sexually transmitted infections indicate a real danger for the rapid transmission of HIV among IDUs and to the wider community.8 Number of known HIV/AIDS cases remains relatively low, but its rise shows signs of acceleration recently. As of December 31, 2007 a total of 1500 HIV cases have been registered in Georgia historically; among them 1142 were males, and 358 – females; 621 patients developed AIDS and 315 persons died. The vast majority of people living with HIV/AIDS were aged 29-40-years at the time of diagnosis. Total number of newly registered cases was 344 in 20079. Georgia’s HIV cases are mostly concentrated among the injecting drug users so far. Among the cumulative HIV cases with a known route of transmission, 60.2% were infected through injecting drug use in 2007. The worst affected areas were Tbilisi, the capital city and Black Sea coastal regions of Georgia (Samegrelo and Adjara). How many IDUs are there in the country? There are presently no valid data on the prevalence and incidence of drug abuse among the general population of Georgia. According to the last available information source, 24,000 people were registered in the country at the end of 2004, 14,400 of them—injecting opioid users. Until 2005, the national database of known / institutionalised drug users (Narcologic Register) was maintained by the Georgian Research Institute on Addiction under the control of the Ministry of Health, Labour and Social Affairs and also sent to the Ministry of Internal Affairs. The Central Information Bank served as the database that contained information on all the registered drug users and registered drug dependent persons across the country.

In May 2005, the database was transferred to the Ministry of Justice’s Bureau of National Expertise (NFB), and later, in November 2006, to the Ministry of Internal Affairs. The NFB and the MOIA are not required to send the information to the Central Information Bank, which consequently lost its active function and is now limited to the “frozen” data collected prior to May of 2005.10 Following the transfer, the database was not updated. No figures of registered drug users are available.

6 Drug Situatuion in Georgia, Annual Report 2005, SCAD 7 Anti-drug Strategy of Georgia, 2007 8 UNGASS Country report Georgia, 2006 9 Infectious Diseases, AIDS and Clinical Immunology Research Center, Annual Report, 2007. Unpublished. 10 The system of registration and follow-up of drug users in Georgia, 2008, David Otiashvili & Nino Balanchivadze, Addiction Research Center Alternative Georgia

US Department of State International Narcotics Control Strategy Reports (2005, 2006, 2007, 2008), released annually by the Bureau of International Narcotics and Law Enforcement Affairs show the drug use trends in Georgia:

2005 – “Independent and official sources indicate that there were at least 275,000 drug users in Georgia during 2004. The increase in the number of drug addicts and drug consumption in comparison with last year’s figure of 150,000 is mainly caused by the import and illegal sale of Subutex. This drug is not registered in the Georgian health care system and is imported illegally mainly from Europe. The price for one tablet of Subutex is approximately $100. The tablet is dissolved into an injectable solution for three or four people at $25-$30 cost per user.”

2006 – “There are no widely accepted figures for drug dependency in Georgia. Press reports indicate at least 350,000 drug users in Georgia during 2005; the government puts the number at 240,000. Any increase in drug consumption is probably due to the growing popularity of Subutex.”

2007 – “There are no widely accepted figures for drug dependency in Georgia, and more generally, statistics are poorly kept. Some sources put the number of drug users between 240,000 and 350,000. Such calculations are, however, at best, a guess. They result from multiplying known users by a coefficient to account for the covert, hidden nature of the problem and poor record keeping.

The GoG has just restarted a national register on drug abusers, which at the end of 2004 numbered 24,000. The register had fallen into disuse after mandatory drug testing was moved from the Ministry of Health to the Ministry of Justice. There were 1488 new registered drug abusers between May-December 2005, with another 4380 registered from January 2006 through mid-October 2006. New figures for 2005 and 2006 are, however, for Tbilisi only. All figures include both hard-core addicts as well as other users.”

2008 – “There are no widely accepted figures for drug dependency in Georgia, and more generally, statistics in this subject area are poorly kept. Some sources put the number of drug users between 240,000 and 350,000. Such calculations are, however, at best, a guess. They result from multiplying known users by a coefficient to account for the covert, hidden nature of the problem and poor record keeping. The Ministry of Justice’s National Forensic Bureau maintains annual statistics on persons tested for drug abuse. In the first 9 months of 2007, the number jumped to 9,581 persons, compared to a total of 5,779 in 2006.”

Another source of information regarding drug situation in Georgia is Drug Annual Reports prepared by the South Caucasus Anti-drug programme (SCAD programme). The South Caucasus Anti-Drug (SCAD) Programme is the response of the United Nations Development Programme (UNDP), the European Union (EU) and national governments of the South Caucasus to reinforce drug control capacities in the region:

“The latest information reflects up to 24 000 drug users registered in the Narcologic Register by the end of 2004. Out of those, up to 14 400 were injected opioid users. To get a relatively accurate description of the situation, the existing officially registered data were usually multiplied by a certain index, differently by different experts (the value of index ranging from 8 to 10). As a result, the number of drug users for 2004 was estimated as 200 000 - 240 000.

The year 2005 showed an increase in the demand for treatment of drug dependent individuals (by 99% for detoxification of opioid-type drugs addicts). This is demonstrated by a sharp increase in the

14 number of treated cases compared to the previous years: namely, 603 patients versus 300 treated in 2004 and 320 treated in 2003. Another interesting trend observed in 2005 was that long queues of patients started to form to register in advance for treatment in the in-patient hospitals, which had never happened before.

In the recent period the death cases caused by drug use are not actually registered by the relevant agencies in the country; to fill in this gap, a special research was implemented in 2005 in the framework of SCAD programme directed on estimation of mortality of drug users. The research revealed, that for the year 2003, mortality of drug user men of reproductive age represented 6 persons per 1000 people, which two times exceeds the mortality rate for the total population of males in 2003 (3 persons per 1000 individuals).”11 Brief overview of size estimation methods It is difficult to overestimate the importance of obtaining accurate information on the prevalence of illicit drug use. Such information is valuable both in terms of monitoring the impact of drug misuse at both national and local levels as well as in assessing the effectiveness of drug prevention efforts.

It should therefore be remembered that, no matter what approach to prevalence estimation is taken, the picture produced by this process can only ever be an imperfect approximation of the real state of affairs. As a result, our knowledge of the world of illicit drug use and our ability to estimate the number of people using illicit drugs within a locality is less complete than we may judge to be desirable.

A variety of methods are available for estimating the prevalence of heavier or more problematic patterns of illegal drug use, for example drug dependence. These include: population-based surveys (although, these are often unreliable for rarer, stigmatized and hidden patterns of drug use); case-finding studies; capture-recapture estimates; multiplier techniques; nomination techniques, including snowball sampling; synthetic estimates, based on social or demographic variables assumed to correlate with drug prevalence; and a variety of more sophisticated statistical modeling approaches.

The above mentioned methods for estimating the prevalence of drug use can be broadened in two wide categories: ⇒ DIRECT METHODS - enumeration (counting) of known drug users and conducting surveys (such as Enumeration of known drug users, population surveys, school-age surveys); ⇒ INDIRECT METHODS - estimating numbers from samples of known drug users (capture- recapture, multiplier method)12.

Direct methods are relatively well equipped for estimation of the overall extent of drug use (of any kind and pattern) in the population; however, they fail to estimate within any acceptable margin of error more rare (low-prevalent) modes and patterns of drug use, such as injection drug use and problem drug use.

Why is it necessary to be concerned with the methodological issues of estimating drug problems? Difficulties in describing the extent, nature and impact of substance use and misuse present considerable scientific challenges. Drug use is usually illicit and hidden and subject to rapidly changing fashions. Routine surveillance sources remain only partially validated, are of changing and, in general, unknown coverage, and measure only a part of the phenomenon. Research

11 Report to the UNDP and EMCDDA by the SCAD National Focal Point, Georgia Drug Situation 2005 12 Approaches to Estimating Drug Prevalence in Ireland: An Overview of Methods and Data Sources, 2003

15 studies are usually conducted in selected populations of unknown representativeness, and with little opportunity for methodological development or collection of time-trend data.

Using multiplier-benchmark methods Of all the methods of indirect estimation the multiplier-benchmark approach is probably the easiest to implement and probably the one with the longest history of use in the field of drug epidemiology. There is a flexibility in how it is applied that makes it useful in many circumstances. In the standard application, it uses information about the known size of an identifiable subsection of the target population of drug users, and generalizes from that subsection to give an estimate of the complete target population by applying a multiplying factor.

In multiplier-benchmark studies, the research makes use of preexisting data for some behaviour or event that is common in the target population of problem drug-taking, for example, police arrest data for drug use or possession, accident and emergency ward data and, more directly, drug treatment data and data on drug-related deaths. Such pre-existing information, which can be simply an anonymous count of the key behaviour over a fixed time period, is called the benchmark information. Along with that national data set is required an estimate of the proportion of the target population who have experienced the event, that is, who have been arrested, who have died etc.; the inverse of that proportion is called the multiplier. Estimating the associated multiplier requires, usually, a small, separate sub-study using nomination technique and again, usually, anonymous records are sufficient.

An early paper by Hartnoll and others (“Estimating the prevalence of opioid dependence”, Lancet, vol. 338 (1985), pp. 203-205) illustrates the application of the simplest technique, using deaths amongst drug users. To apply the multiplier procedure to estimate the number of drug users in a given year, he uses two things: ⇒ The number of deaths to drug users in that year, say 3,000; that acts as the fixed benchmark in the calculation; ⇒ The death rate amongst drug users in that year, say 2 per cent, or 1 in 50 dying in the year; that provides the multiplier in the calculation.

The estimate of the number of drug users in that year is calculated from those two figures as the population size required for a 2 per cent death rate to result in 3,000 deaths. If 1 in 50 die, then the overall population must have been estimated for approximately 3,000 x 50 = 150,000. The calculation is notable for its simplicity and directness.

Relative trends inferred from routine information systems and agency data Before looking at estimation methods, it is useful to consider the uses of data from routine information systems. Although such sources of data of themselves do not provide the actual number of drug injectors in a population, they are often used in providing relative trend data.

An example of routine information systems could be data from drug treatment agencies. This kind of data is often considered to be reflective of the larger unknown population. For example, it is sometimes assumed that if the number of injectors coming to treatment has increased, then this reflects an increase in the number of injectors in the population – other things being equal.

Trends can often be inferred from existing sources such as: ⇒ data from health centres and treatment clinics - including characteristics of drug injectors such as age, sex, type of drugs used, route of drug administration, and prevalence of hepatitis, and HIV/AIDS ⇒ data from enforcement agencies - such as the range of available drugs, their purity, street prices, drug trafficking routes, and localities of drug use

⇒ data from hospitals and emergency units: such as the number of cases treated, trends in infectious conditions such as hepatitis, HIV/AIDS, and the number of reported overdoses ⇒ data from national health surveillance systems and disease registers: such as the incidence and prevalence of hepatitis, and HIV/AIDS

The proportion of the target population in the benchmark may be obtained separately and independently by interview/questioning or by other specific studies. Sometimes it is possible to use figures from already published data, if they are appropriate for the target population, or even from a general population survey itself, if it contains a high number of drug user respondents from the target population. There are a range of different types of multiplier study that can be carried out, including nomination studies.

Nomination techniques The use of nomination methods as a means of obtaining information about difficult to-reach populations dates back many years having enjoyed a certain amount of fame and notoriety in the 1970s. Interest in these methods is now developing again in drug use epidemiology, its main virtue being its usefulness in dealing with relatively rare events. Nomination techniques are estimation methods based on information which individuals in a sample provide about their acquaintances. It is similar to the multiplier technique, and prevalence is estimated using the benchmark/multiplier approach. It differs in that it gets its multiplier from information gained from informants who are interviewed.

Broadly put, sample members are asked to name or nominate drug-using acquaintances and to say whether these acquaintances have been in touch with drug treatment centres, health services or any other similar body, within a stipulated time period. The proportion of treatment attendees nominated by the sample is then used as a multiplier as described above, in conjunction with the benchmark of known attendance figures at the drug treatment agencies, to give an estimate of the total number of drug users.

There are four steps in using a multiplier/benchmark method to estimate the prevalence of behavior or characteristic amongst the total population: ⇒ Select a benchmark where data are available and you are confident in the data provider, or ask the appropriate questions during the rapid assessment. ⇒ Select a multiplier - using data from research studies (It is recommended that, whenever possible, the researcher should conduct a sample survey of the target population - injectors or problem drug users - as part of the prevalence estimation study, e.g. survey using nomination technique). ⇒ Calculate the number of cases by multiplying the benchmark by the multiplier. This will give the estimated number of cases. ⇒ A further step can be an estimate of the absolute prevalence of that behaviour or characteristic amongst the total population. Respondentdriven Sampling (RDS) as a Strategy to Reach Hardtoreach Populations The problem of collecting accurate information about the behavior and composition of social groups arises in many areas of research. In most cases, standard sampling and estimation techniques, developed over the past years, provide a means for collecting such information. However, there are a number of important groups for which these techniques are not applicable. Traditional probability- based sampling methods require the development of a sampling frame, which is challenging for hard-to-reach or “hidden” populations (Robinson et al. 2006).

To minimize selection bias, researchers have started using a new sampling alternative among populations such as MSM, commercial sex workers, and injection drug users. This sampling method, respondent driven sampling (RDS) (Heckathorn 1997, 2002), is a type of chain-referral sampling, or snowball sampling, and “is based on the recognition that peers are better at locating and recruiting other members of a hidden population than outreach workers and researchers” (Semaan et al. 2002).

This method combines “snowball sampling” (getting individuals to refer those they know, who in turn refer those they know and so on) with a mathematical model that weights the sample to compensate for collecting it in a non-random way. A dual compensation system, whereby a respondent is compensated for participating in the study and for recruiting his/her peers, is used. Moreover, proponents of RDS claim that this sampling method can produce probability samples of the target population and reduce several sources of bias found in chain referral methods (Heckathorn 1997, 2002; Semaan et al. 2002). This approach reduces bias associated with the choice of initial participants, volunteerism and masking by using steering incentives for participation and recruiting participants. Thus, as a result of the successive waves of recruitment, it does not matter whether the initial sample is randomly drawn. The population-based estimates are based on a model that takes into account the network size of participants and recruitment patterns (Heckathorn 1997; Salganik and Heckathorn 2004; Thompson and Frank 2000).

There are several advantages to respondent-driven sampling. Among the primary features that distinguish RDS from snowball sampling is that ‘seeds’ are limited in the number of respondents they can recruit by the number of coupons they receive (e.g. three to four), thereby minimizing the influence of initial seeds on the final sample composition. Limiting the number of recruits in this way encourages long recruitment chains, thereby increasing the ‘reach’ of the sample into more hidden pockets of the population (Magnani et al. 2005). Next, the sample provides information about the people in the population and the network connecting them. Another desirable property is that sample data can be combined with institutional data to estimate the size of a hidden population. Previous methods for estimating the sizes of hidden populations did not allow for unbiased estimates of population composition. Respondent-driven sampling is also cheaper, quicker and easier to implement than other methods commonly used to study hidden populations. Complementary Projects Since 2008, program entitled “Establishment of Evidence-based Basis for HIV/AIDS National Program by Strengthening Surveillance System” is being implementing within the framework of the Global Fund Project to Fight AIDS, Tuberculosis and Malaria (GFATM) by Curatio international Foundation (CIF) in cooperation with AIDS and Clinical Immunology Research Center, National Center for Disease Control and Public Health (NCDC) and local NGOs Bemoni and Tanadgoma. Under this project Bemoni proposed to conduct Behavioral Surveillance Surveys (BSSs) with a Biomarker Component among injecting drug users (IDUs) in five main cities of Georgia: Tbilisi, Gori, Telavi, Zugdidi, and Batumi. It was decided to use this opportunity and incorporate the nomination study for estimating the size of the injecting drug user (IDU) population into the above mentioned BSSs. Under the framework of the Global Fund Project the main field work activities would be covered, this could allow saving time as well as financial and human resources for present study.

To avoid duplication of efforts and to promote complementary activities, discussions were held between the SCAD and the GFATM representatives, and subsequently with CIF, to harmonize objectives and research plans. As a result of these discussions, informal collaboration of the two projects was agreed.

METHODOLOGY Objective of the Study The aim of the study is to estimate the prevalence of Injection Drug Use (IDU) in Tbilisi (the capital) and 4 main cities (Batumi, Telavi, Gori and Zugdidi) of Georgia and provide IDU prevalence estimate throughout the Country.

Objectives: 1. to undertake population estimation studies using consistent methodologies; 2. to recommend methods for use in other sites across Georgia; 3. to provide an evidence base of estimates of the prevalence of injection drug use in Georgia; 4. To help establish a monitoring system that will track injection drug use trend data Defining the Target Population Problem Drug Use (PDU) is defined as injecting drug use or long term/regular use of opiates and/or cocaine-type drugs and/or amphetamine-type drugs13. Taking into the consideration the fact that within the framework of the Global Fund Project to Fight AIDS, Tuberculosis and Malaria BPU intends to conduct Behavioral Surveillance Surveys (BSSs) with a Biomarker Component among injecting drug users (IDUs) and we have the opportunity to incorporate the Study of estimating the size of the injecting drug user (IDU) population into the above mentioned BSSs, it should be mentioned that in this report we imply only injection drug users.

Thus, for the purpose of this study, we regarded any person who has used any psychoactive drug through injections in a non-medical context.

Inclusion criteria - to be eligible, each participant must meet the following criteria: 1. Aged 18 years or older 2. Lives in the participating city/district 3. Has not previously completed an interview under the current study 4. Able to complete the interview in Georgian 5. Arrives at the study site with a valid study recruitment coupon. 6. Currently injects drugs (this was identified by reported drug injection in the month prior the survey) 7. Has either: ⇒ Physical evidence of recent injection (fresh track marks, scabs, or abscesses), OR ⇒ Knowledge of drug prices, preparation, injection, and etc. Stages of IDU prevalence estimation method to be applied for Georgia There are five stages of prevalence estimation method that had been used in this study.

Stage 1: Data collection of IDUs (gaining the benchmark data - B) All available data on injection drug use in Georgia were reviewed. Data of IDUs are recorded under the current system for the year 2007 (details see below in chapter “Benchmark Data Collection”).

Stage 2: Estimation of the value of multiplier (M) The proportion of the target population in the benchmarks is obtained from research studies using nomination techniques (study using the Respondent Driven Sampling (RDS) methodology based on appropriate eligibility criteria and accurate sample size calculations was conducted). The survey

13 EMCDDA Recommended Draft technical Tools and Guidelines. Key Epidemiological Indicator: Prevalence of Problem Drug Use. EMCDDA/ July 2004 19 collected the data among IDUs using nomination method/questionnaire developed by SCAD epidemiology experts.

Stage 3: The derivation of multiplier - this stage involves two steps: a) Estimation of the percentage (P) of IDUs recorded from Stage 2. Separate estimates for different benchmarks were made in each city - percentage of IDUs tested by police for presence of illegal drugs in 2007; % of IDUS tested for HIV in 2007; % of IDUs considering entering the abstinence-oriented treatment in 2007; % of IDUs in substitution treatment in 2007; % of IDUs in substitution treatment waiting list in 2007; % of IDUs in the needle exchange and other low- threshold programs in 2007; % of IDUs deceased due to a fatal drug overdose in 2007. b) Multiplier (M) is estimated for each benchmark by the inverse of percentages (Pisani, 2002). The formula M = 100/P

Stage 4: Estimate the number of drug injectors Numbers of IDUs estimates for each benchmark are obtained by multiplying the recorded number of IDUs (collected from the available data source) by an appropriate multiplier (The formula E = BxM). For example, if this method is applied to in-treatment data then the benchmark is the total number of drug-users who underwent treatment in a given year, the multiplier is the in-treatment- rate (the proportion of treatment attendees nominated by the sample). Those two components—the known figure in treatment contact (the treatment benchmark) and the estimated proportion of abusers who were in treatment contact (giving the treatment multiplier) - are what gives the method its name.

Case study. Toronto multiplier study based on HIV tests (Basic multiplier-benchmark calculation)14 Archibald and others (2001) outlined a multiplier method of estimating the prevalence of injecting drug use, making use of information from laboratories of the number of HIV tests by injecting drug users and of data from surveys of the proportion of injecting drug users that had had an HIV test in a given year. The findings for one city in one year, Toronto in 1996, are presented below.

The example requires two elements. The first is a known benchmark figure. That figure, in the present case, is the number of HIV tests made on injecting drug users in Toronto in 1996, which was recorded in routinely collected information as 4,050. That represents the known part of the population of injectors. to find the total number of injectors, it needs to be determined what fraction of them are unknown to HIV testing records. The second element required by the method is therefore a multiplier that tells how many more injecting drug users in Toronto did not have HIV tests in 1996. That figure can be worked out simply if the proportion of drug users who did have HIV tests during the period is determined. In the example, the proportion of users tested for HIV was known from other studies to be 25 per cent, or 1 in 4. The calculation illustrated below (Table 1.) in is then made simply by noting that if 1 in 4 injectors have been tested, then the total number of injectors must be 4 x 4,050, or 16,200, people.

Table 1. Basic multiplier-benchmark calculation

Item Applied Values Estimates Benchmark (B) Number of HIV tests by injecting drug users in 1996 4050

Proportion of injectors reporting getting an HIV test in the 25 per cent previous year (P)

Multiplier (M) Multiplier calculated as 1.0/0.25 (i.e., 1 in 4)15 4,0

Population estimate Benchmark times multiplier (B*M) 16 200

14 Estimating Prevalence: Indirect Methods for Estimating the Size of the Drug Problem, Global Assessment Programme on Drug Abuse. UNODC, Vienna, 2003

15 This is the same as 100/25 (M = 100/P) 20

Stage 5: Calculation of a prevalence of drug injection for each selected city It was based on data on population distribution (State Department of Statistics of the Ministry of Economic Development of Georgia). Census data gave the population for urban areas. The population between 18 and 64 was used as the denominator for the prevalence based estimate. The appropriate estimates of injecting drug use were then applied to that adult population. An upper and lower limit is provided by statistical means.

Additionally, the first attempt to derive the national estimate for the percentage of injection drug users in Georgia using the Multiple Indicator Method (MIM) had been carried out. Limitation of the study No matter what method is used, all data are potentially biased for a variety of reasons. The multiplier methods is relatively straightforward to use, but will depend on good institutional record- keeping. The greatest difficulty in using multiplier methods correctly is finding data from institutions and populations that correspond with one another. To use institutional and survey data together to estimate the size of a population, the members of the population all have to have a chance of being included in both the survey and in the institutional data (for example because they have access to that service).

Multiplier methods using treatment, police, or mortality data are ad hoc methods. They are not based on statistical theory and no formula for the variation of the estimator can be derived. Benchmarks are usually collated on a national level. The corresponding multipliers are derived from local samples or expert ratings. Their validity for the total population is questionable due to regional and temporal variations. These methods are easy to apply and give only point estimates.16

Sources of information used for estimations may limit the generalisability of the final estimates. Here are some examples of how this happens: ⇒ Drug treatment programs typically attract chronic, long term IDUs at the conclusion of their drug using careers, under-representing newer drug users. ⇒ Jails and criminal justice settings will have fewer newer IDUs under-representing long-term users and those not involved in criminal activities to support their drug use. ⇒ Clinic settings will under-represent healthier drug users. ⇒ Methadone treatment programs will only yield information about opioid users, private programs will only include IDUs that can afford to be in treatment. ⇒ Low threshold agencies may collate the same standard of information on their clients as the more formal drug treatment agencies described above, and some clients may only be known by a forename or an assumed name.

Depending on the point of contact sources used, we may have to adjust estimates to reflect their relationship to a wider population of interest. It will be best to use as many sources as we can. City-wide service points of contacts or institutional data that are widely accessed by IDUs and covers the highest numbers (and types) of IDUs should be used for gaining benchmark data. Decent data are often just partly available or not at all, but, once the importance of collecting reliable data is recognized on the political level this problem can be solved. Then, by means of good sampling, sound survey instruments and by means of good police registration techniques, prevalence and patterns (of different kinds) of drug use can be studied on a regular basis. And once treatment institutions are in place, reliable treatment registration can supply interesting data sets as well.

16 Study to Obtain Comparable National Estimates of Problem Drug Use Prevalence for all EU Member States, Final Report. EMCDDA Project (CT.97.EP.04)

The prevalence estimation obtained in this study should be treated with caution, as the data are self-reported; underreporting or over-reporting of behaviors is possible yet difficult to ascertain. Another issue is that ideally multiple benchmark data sources (and hence a variety of multipliers) should be used in a prevalence estimation exercise. Unfortunately different numbers of benchmarks are available in different cities of Georgia. For example, only police and HIV testing data are available in Telavi.

RDS Possible limitations to the study could have affected the results. The small numbers of women participating in the surveillance may indicate a strong desire to remain hidden, their limited numbers, or a reflection of poor recruiting.

Another study limitation for IDUs was related to the inclusion criteria adopted. Due to the need of parental consent for enrollment of those aged 15-17, this age group was not represented in the sample, especially in light of the fact that the results showed that 56.2% of survey participants in Tbilisi started injecting drugs in age 15-19. Study Design

Selection of sites Selection of data sources Questionnaire development Training of staff Seeds were recruited And interviewed

R Each seed referred 3 known IDUs D Recruits were interviewed S and asked to refer 3 more IDUs

Data from total sample of Collection the data IDUs analyzed from routine data sources

Multiplier Data from both sources used to generate estimates Method

Estimated prevalence of PDU in selected areas

Figure 2. Study Design

Prestudy activities Negotiating access to data sources It is helpful when beginning a research study to have a very clear idea of what data and information sources are routinely available and which of those can be accessed for extracting information relevant to the study. A drug misuse prevalence study can only be undertaken with the co-operation of those who hold information on drug misuse. Each agency will have its own idea about the need or relevance of prevalence research, and each agency will have its own concerns about giving access to confidential data. Agencies which are not exclusively concerned with drug misuse may see requests for information on drug misuse as an additional burden which they may not be keen to take on. They may also be more political obstacles to collecting data from some agencies. The main issue which agencies see as a reason for not giving access to their data is confidentiality.

Expert Team Leader selected appropriate experts and established expert team in order to derive multipliers for predefined drug-using sub-population groups (the benchmarks). The experts had been hired from those institutions where data are available and Project team was confident in the data provider (Deputy Director of the Research Institute on Addiction, Director of the Drug Prevention Center within the National Center for Disease Control and Public Health (NCDC), representatives from the AIDS Center and Ministry of Internal Affairs).

Development of the nomination questionnaire Nomination questionnaire was developed by SCAD programme epidemiological experts. The questionnaire had been translated and pre-tested with members of the target population. In this regard 3 pilot focus group discussions (FGD) were conducted in October 2008. The first two focus group discussions were attended by six male drug users each, and 7 male drug users participated in the third focus group. The initial version of the questionnaire was tested in the first and the second groups, and then updated and modified in view of the recommendations and comments of the group participants. The second version of the questionnaire, i.e. the one updated after the first two groups, was tested in the third group.

Training of the survey staff Before the implementation of the study, the staff received 1-day training in study procedures, instruments, and interviewing process, conducted by Project Coordinator and SCAD Epidemiology Expert. Participatory nature of the training program ensured that all staff members understood their respective roles and responsibilities fully.

Geographical Scope As it was mentioned above within the framework of the Global Fund Project to Fight AIDS, Tuberculosis and Malaria Bemoni conducted Behavioral Surveillance Surveys (BSSs) with a Biomarker Component among injecting drug users (IDUs) in five main cities of Georgia: Tbilisi, Gori, Telavi, Zugdidi, and Batumi. The Study of estimating the size of the injecting drug user (IDU) population had been incorporated into these BSSs. The map below (Figure 2.) shows the cities where survey was conducted.

Figure 3. Map of Georgia with Study Sites Study sites and staff The interviewing process took place at the: ⇒ Bemoni office in Tbilisi (November, 2008) ⇒ Station of blood transfusion in Gori (December, 2008) ⇒ Needle exchange Center in Telavi (February, 2009) ⇒ NGO “Tanadgoma” brach in Zugdidi (March, 2009) ⇒ NGO “Tanadgoma” brach in Batumi (April, 2009)

All sites were accessible to study participants by public transport. Participant flow at the study sites were designed to provide maximum privacy and minimize their exposure to any other study participants. The sites were open from 10am to 8pm (Monday to Saturday). Each office was supervised by the study coordinator and staffed at all times by field coordinator, addiction specialist, 3 interviewers, coupon manager, and social worker. Nearly all of them had previous experience working on similar research projects in the recent past. Ethical Considerations The study investigators are cognizant of the fact that the target groups for this study are at some risk for social harm should they be identified as part of the target groups. We have designed this study to maximally protect the participants balanced with the individual benefit and community benefits from this study. Specifically, ⇒ Initial identification of areas where sampling took place was done by NGOs currently working with and trusted by the populations. ⇒ Informed consent was taken in a staged manner. ⇒ No names had been recorded. All documentation is anonymous.

Given that parental consent is required in Georgia for individual below the age of 18, we did not recruit participants below this age for the survey.

Participation of all respondents in BSS and Estimating the Prevalence of Problem Drug Use surveys is strictly voluntary. Measures were taken to assure the respect, dignity and freedom of each individual participating. During the survey emphasis was placed on the importance of obtaining informed consent (orally), and avoiding coercion of any kind. Complete confidentiality of study subjects was also emphasized. Names of respondents were not be recorded anywhere on the questionnaires or other forms. Study documents, including RDS data and blood specimens, were identified using unique ID numbers for each participant to maintain confidentiality. Steps in Data Collection RDS Recruitment of respondents was conducted using RDS. Based on sample size calculation in the framework of Behavior Surveillance Surveys sample size was defined as 300 IDUs in Tbilisi and 200 IDUs in each other selected city (Gori, Telavi, Zugdidi and Batumi).

Sampling Procedure 1. The process starts with recruiting initial participants who are considered as ‘seeds’, who were selected non-randomly. The criteria for seed recruitment are: those who have different sociodemographic characteristics, at different locations, who have awareness of networks of target populations.

2. Selected seeds underwent eligibility checking: In order to ensure that authentic IDUs are recruited and not just individuals wanting money, a verification procedure was done by the experienced addiction specialist (narcologist). This verification procedure included a preliminary informal discussion regarding the street names of drugs and prices, familiarity with drug preparation and injection techniques and finally visual inspection for recent track marks. If the narcologist was satisfied with the recruit’s responses, the interview was conducted.

3. After the eligibility check witnessed verbal informed consent for the interview was obtained (for confidentiality reasons, including legal and moral undertones, it is recommended that informed consent should only be elicited verbally) - those who were eligible and willing to participate in the study had to go through the informed consent procedures in a private area: the participant received information about the studies as well as the informed consent procedure and was asked to accept willingness of participation; after that 2 staff members signed the informed consent form on behalf of participant. Additional information was collected that was specifically required for RDS methodology: personal network size, relationships to recruiters, and the number of recruitment refusals encountered. Following the informed consent process, the field coordinator administered a face-to-face interview with the participant about the participant’s personal network. After that participants completed the Interview. All interviews took place in private rooms with only the interviewer and subject present.

4. Interviewed seeds were given an incentive (20 GEL) for participation in the study. Once initial participants completed their interview each seed received three unique, non-replicable, recruitment coupons with a two-week expiration date to recruit their peers who also fit the eligibility criteria for the study. These peers are no longer considered ‘participants’ and are referred to as recruiters. (When a study participant is recruited by a recruiter, but has not yet enrolled in the study, that person is referred to as a ‘recruit’.)

5. Seeds were offered incentives to recruit their peers into the same interview they have just completed; the recruiter was given 21 GEL for three recruited peers. The first wave of participants recruited for the study was brought in by ‘seeds’. Thereafter, each person recruited for and enrolled in the study received personal ID and three recruitment coupons with which to recruit their peers into the study as well. Recruits should have to present for participation with coupon “in-hand”. The limitation of three peers per recruiter was done to ensure that a broad 25

array of subjects have an opportunity to recruit. Respondents received compensation for participating in the study and for each of their recruits who subsequently enrolled in the study.

6. Each coupon is uniquely coded in order to link recruiters with recruits. Personal ID as well as the coupon ID numbers were carefully recorded in each questionnaire. Coupon numbers (received and given) become part of the information entered into the computer record for each respondent. Every recruit who visited study centers were marked on spreadsheets for coupon management.

7. All new recruits were offered the same dual incentives, as were the seeds. Everyone had been rewarded both for completing the interview and for recruiting his or her peers into the survey. We will perform this recruitment for six waves total or until the sample size is reached.

8. Three coupons continued to be distributed until sample sizes were attained, after which participants were warned that the study would be ending within a few days. However, participants were informed from the beginning that once sample sizes were reached, no more coupons would be honored.

1 2 3

Give three coupons to Identify, recruit, and Inform seeds on who and each seed to recruit interview seeds. how to recruit. IDUs in their network.

4 5

Recruits bring valid coupons Participants are offered the to the study site. If eligible, chance to recruit others. Those they are interviewed. who agree are informed and given three coupons.

Seeds/recruiters are rewarded for every person enrolled and interviewed.

Figure 4. RDS Recruitment Methods Sample Sizes Participating organizations (Bemoni in Tbilisi and local service provider organizations in other cities) working with IDUs recruit 4 to 7 IDUs to serve as “seeds” (7 “seeds” in Tbilisi, 5 in Gori and Telavi, 4 in Zugdidi and 6 in Batumi had been recruited). All the 27 seeds were productive. One seed recruited one participant, four recruited two, and 2 recruited three participants in Tbilisi; two seeds recruited one participant and three recruited 3 participants in Gori; and in other cities all seeds recruited 3 participants. From these initial recruited participants, the recruitment chains were very different, as observed in Figure ?. The chains of these seeds are the first persons of the target

26 group to be contacted and the first to recruit peers and refer for interview. The diverse characteristics of the seeds are shown in Tables 2-6.

Table 2. Socio-demographic characteristics of seeds vs. total sample in Tbilisi

Seeds Sample

Tbilisi RDSAT adjusted n= 7 % n=307 SPSS (%) (%) Age 18-24 1 14.3 21 6.8 7.6 25-30 1 14.3 51 16.6 16.8 31-40 2 28.6 80 26.1 26.5 41-50 2 28.6 122 39.7 38.8 50+ 1 14.3 33 10.7 10.3 Gender Male 6 85.7 304 99 99,3 Female 1 14.3 3 1 0.7 Marital Status Married 3 42.9 167 54.4 53.6 Divorced 1 14.3 62 20.2 21 Has never been 3 42.9 25.4 78 25.4 married Education Level 27.1 Secondary 82 26.7 1 1.3 Incomplete High 16 5.2 5.3 High 6 85.7 209 68.1 67.6

Table 3. Socio-demographic characteristics of seeds vs. total sample in Gori

Seeds Sample

RDSAT adjusted Gori n= 5 % n=205 SPSS (%) (%) Age 18-24 35 17.1 23.1 25-30 38 18.5 20.0 31-40 1 20 74 36.1 33.6 41-50 4 80 50 24.4 18.4 50+ 8 3.9 4.9 Gender Male 5 100 200 97.6 97.7 Female 5 2.4 2.3 Marital Status Married 3 60 111 54.1 54.6 Divorced 1 20 20 9.8 9.2 Has never been 1 20 36.2 74 36.1 married Education Level Secondary 3 60 132 67.3 68.0 Incomplete High 1 20 73.4 3 High 1 20 60 29.3 28.9 27

Table 4. Socio-demographic characteristics of seeds vs. total sample in Telavi

Seeds Sample

Telavi RDSAT adjusted n= 5 % n=205 SPSS (%) (%) Age 18-24 1 20 34 16.6 18.3 25-30 1 20 56 27.3 29.5 31-40 3 60 75 36.6 35.8 41-50 36 17.6 14.9 50+ 4 2.0 1.5 Gender Male 5 100 205 100 100 Female Marital Status Married 2 40 85 41.5 41.7 Divorced 21 10.2 10.4 Has never been 3 60 47.9 99 48.3 married Education Level Secondary school 4 80 139 67.8 66.3 Incomplete Higher 5 2.4 3.8 Higher 1 20 61 29.8 30

Table 5. Socio-demographic characteristics of seeds vs. total sample in Zugdidi

Seeds Sample

Zugdidi RDSAT adjusted n= 4 % n=204 SPSS (%) (%) Age 18-24 27 13.2 15.4 25-30 1 25 45 22.1 22.6 31-40 78 38.2 37.9 41-50 40 19.6 18 50+ 3 75 14 6.9 6.1 Gender Male 4 100 203 99.5 99.5 Female 1 0.5 0.5 Marital Status Married 3 75 106 52.0 50.3 Divorced 8 3.9 4.6 Has never been 1 25 45.2 90 44.1 married Education Level Secondary 3 75 118 57.8 57.4 Incomplete High 12 5.9 6 High 1 25 74 36.3 36.6

Table 6. Socio-demographic characteristics of seeds vs. total sample in Batumi

Seeds Sample

Batumi RDSAT adjusted n= 6 % n=206 SPSS (%) (%) Age 18-24 1 16.7 25 12.1 12.5 25-30 2 33.3 47 22.8 24.6 31-40 2 33.3 78 37.9 37.5 41-50 1 16.7 48 23.3 22.1 50+ 8 3.9 3.3 Gender Male 5 83.3 200 97 98.1 Female 1 16.7 6 3 1.9 Marital Status Married 1 16.7 99 48.1 49.1 Divorced 2 33.3 27 13.1 12.4 Has never been married 3 50 80 38.8 38.5 Education Level Secondary 2 33.3 142 68.9 70.2 Incomplete High 1 16.7 15 7.3 7 High 3 50 49 23.8 22.8

Overall, 1188 IDUs were recruited (including seeds) between November 10, 2008 and April 4, 2009 (the illustrative sample in Tbilisi, using NetDraw, a network illustration program, is presented below in Figure 4). Average duration of recruitment process in each site was 14 days. A total of 3064 coupons (865 in Tbilisi, 561 in Gori, 615 in Telavi, 510 in Zugdidi and 513 in Batumi) were handed out to participants to recruit their peers. Of the 1188, 5% (59) were ineligible to participate in the study for reasons listed in Table 7. Two eligible participants refused to participate in BSS. Thus, 1127 eligible IDUs participated in the BSS study.

Figure 5. Sample of Recruitment Pattern in the Study (Tbilisi)

Legend for Sample Age HIV positive – Black Circle – 18-24, Square – 25-30, Triangle – 31-40, Crossed square – 41-50, Turned triangle -50+ Seeds - green

87 (12 in Tbilisi, 14 in Gori, 26 in Telavi and Zugdidi and 9 – in Batumi) refused to answer the questions after administering the nomination questionnaire and dropped out of the multiplier/benchmark study using, leaving a total of 1040 participants. The reasons for refusal were different, some of them did not like to say anything about their friends, some of them said that they injected alone, so they did not have information about other IDUs; others stated that in 2007 they were imprisoned or were out of the country, so did not have real and correct information about their friends. Another reason for refusal was the fact that respondents started to inject drugs only 6 months ago, so they could not answer any questions about 2007 year time period.

Table 7. Ineligibility, refusals and drop-outs

Tbilisi Gori Telavi Zugdidi Batumi # % # % # % # % # % Persons arriving 333 100 210 100 218 100 212 100 215 10 at the study site Ineligible to 25 7,5 5 2,4 13 6,0 7 3,3 9 4,2 participate Reasons for ineligibility Was not an IDU 18 5 13 4 4 Did not inject 5 drugs within the 3 4 past month Was not 18 years 1 old Had already been 1 interwieved Other 1

Refusal (did not 1 1 want to talk about himself) Enrollments in 307 92,2 205 97,6 205 94,0 204 96,2 206 95,8 BSS

Drop-outs 12 3.9 14 6,8 26 12,7 26 12,7 9 4,4 (refused to answer the questions after administering the nomination questionnaire) Reasons for refusal Injected alone 5 3 4 2 1 Was imprisoned 2 3 4 1 Was outside of the 1 4 9 14 2 country did not like to say 6 3 3 4 2 anything about their friends Started injection 2 7 2 3 later Enrollments in 295 96,1 191 93,2 179 87,3 178 87,3 197 95,6 studies using nomination techniques

Socio-demographic characteristics of non-respondents in nomination survey see in appendix 1 (Table 31).

Interview of Respondents Data collection for BSS (under the Global Fund Project) consisted of an interviewer-administered structured questionnaire and a blood sample collection to test for syphilis and HIV infections among IDUs recruited into the study. The BSS core questions assess the participant’s demographic information, drug use history, drug and sex-related risky behaviors, HIV/STI knowledge, opinion and attitudes, HIV testing history, and access to and use of HIV prevention services.

After the eligibility check and informed consent procedures, the questionnaire with unique ID number was assigned and the subjects were brought to interview rooms designed to maintain privacy. Face to-face and individual interviews were conducted in the interview rooms by trained interviewers. Each interview took about 20-30 minutes.

After that nomination questionnaire (with the same ID number) was administered to the respondents (see appendix 3). The average duration of the interview process was 10-15 minutes. The study participants were asked to nominate up to 10 close friends with whom they had been using drugs in 2007 (Table 8).

Table 8. Number of acquaintances nominated by IDUs

N Tbilisi Gori Telavi Zugdidi Batumi

Frequ % Freque % Frequ % Frequ % Frequ % ency ncy ency ency ency 1 6 2,0 5 2,4 8 3,9 7 3,4 6 2,9 2 20 6,5 19 9,3 16 7,8 8 3,9 41 19,9 3 46 15,0 36 17,6 51 24,9 42 20,6 68 33,0 4 57 18,6 44 21,5 53 25,9 49 24,0 26 12,6 5 81 26,4 40 19,5 27 13,2 45 22,1 33 16,0 6 32 10,4 19 9,3 9 4,4 7 3,4 11 5,3 7 18 5,9 15 7,3 5 2,4 6 2,9 5 2,4 8 7 2,3 6 2,9 4 2,0 4 2,0 4 1,9 9 3 1,0 2 1,0 1 0,5 0 0 10 25 8,1 5 2,4 5 2,4 10 4,9 3 1,5 Total 295 96,1 191 93,2 179 87,3 178 87,3 197 95,6 Missing 12 3,9 14 6,8 26 12,7 26 12,7 9 4,4 System Total 307 100 205 100 205 100 204 100 206 100

The respondents were then asked how many of their closest friends had received treatment or been tested by police for drug presence during this time period. Identifier information for nominated peers was used to establish an unknown to known ratio for each site.

When both interviews were completed, participants were guided to rooms for the collection of biologic samples. The blood test was anonymous-linked. ID numbers were used to label containers of biological specimens (blood). Blood specimens were sent to the Laboratory of Serology and Virology of the AIDS Center in Tbilisi for testing and the results were reported back to the organization. The IDUs were asked to return with their identification card and their results would be provided.

Detailed checking of the completed questionnaires was initiated by the study coordinator during the fieldwork. Care was taken to check errors and inconsistencies to avoid any difficulty at the stage of data analysis. By the end of each day, coupon manager entered recruitment data to the Coupon tracking form version 3.117

Benchmark Data Collection Routine statistics have the advantage that they are readily available. If they have been collected consistently, then they can provide indirect indicators of trends over the years. However, they often provide only basic, aggregated information on a small number of variables. A more important limitation is that information systems which are not specifically concerned with drug use are very likely to under-record drug-related cases.

The benchmark data for this study were collected from the following accessible data sources: 1. National Center for Disease Control and Public Health database of IDUs gathers records from different abstinence oriented treatment facilities (Research Institute on Addiction, addiction center in Batumi and the licensed private treatment centers; available in Tbilisi and Batumi This database obtains anonymous data on individuals who are in contact with a range of drug services. The number of centers involved in treatment of drug addicts in 2007 was 7 (five of those were located in Tbilisi and 2 – in Batumi). Total number of beds in these clinics was 45 (32 in Tbilisi and 13 – in Batumi). Medical treatment of drug dependent individuals had not been financed by the State, donor or private organizations in 2007. The patients had to pay for themselves. The treatment was quite expensive. Many individuals, willing to undergo treatment, could not afford it. Thus, the number of treated cases does not reflect the actual level of demand for treatment in the country.

2. Information from the treatment facilities on waiting lists for the abstinence oriented treatment in 2007; available in Tbilisi and Batumi This information had been gathered from the heads of detoxification clinics in Tbilisi and Batumi, based on personal communications.

3. Methadone Program database of attending IDUs; these data also include information on IDUs in waiting lists; available in Tbilisi and Batumi In December 2005, the first Methadone substitution therapy programme was launched in the country. This programme, financed by Global Fund, is coordinated by the Georgian Research Institute on Addiction. 60 patients were selected for treatment at the first stage. In 2006, within the framework of Global Fund’s programme, another centre for substitution therapy had been established in Tbilisi. By that time, all the HIV infected active opioid users residing in Tbilisi that expressed their willingness to undergo treatment, had been included in the substitution programme. In February 2007 the third center started operating in Batumi.

4. Ministry of Internal Affairs database of IDUs The data on Injection drug users come into contact with the police throughout the country is available by special request from the MoIA. Under Article 4518 of the Administrative Code of Georgia, purchase and possession of drugs in minor quantities or use of drugs without

17 This excel file was created for the purpose of assisting the RDS research study in Zagreb, Croatia. Author: Hrvoje Fuchek 18 Article 45 of the Administrative Code of Georgia - “Illegal production, purchase, storage, use without doctor’s prescription of small amounts of psycho-active substances under control in Georgia for individual use”

medical prescription is punishable with fine, or administrative detention. Article 27319 of the Criminal Code of Georgia stipulates that drug use is only qualified as a criminal offence if a person previously subjected to administrative punishment for drug use continues to use drugs without medical prescription during one year following the penalty. Georgian drug legislation does not distinguish between being detained in connection with the use of drugs and being convicted for purchase or possession of drugs.

Information relating to the use of injection drugs is available from the Department of Information and Analysis of MoIA. According to Article 45 of the Code of Administrative Offences, in case of considerable doubt that a person is under the influence of drugs and/or psychotropic substances, or has used drugs, the police officer is authorized to demand that the person in question undergo an examination. A clinical laboratory and/or laboratory test determining the fact of drug use and/or drug and/or psychotropic intoxication is carried out based on the official referral from an authorized police officer. Ministry of Internal Affairs, specifically, the Department of Information and Analysis records all cases where the fact of drug use without appropriate medical purposes has been established.

5. AIDS and Clinical Immunology Research Center database for HIV testing with IDU identifier (form #23) Georgian AIDS and Clinical Immunology Research Center is a main institution responsible for development, implementation and coordination of all activities against HIV/AIDS epidemic spread in Georgia. It is the governmental institution affiliated with Department of Pubic Health of the Ministry of Labor, Health and Social Affairs. The Center was established in 1989 on the bases of the first HIV/AIDS Diagnostic Laboratory at the Hospital of Infectious Diseases in Tbilisi, Georgia.

The Center and its branch offices in various districts and regions (56 Diagnostic Labs) of the country serve entire population of Georgia providing counseling, testing, monitoring and treatment of HIV/AIDS. Also the Georgian AIDS and Clinical Immunology Center coordinates all activities concerning of HIV/AIDS prevention in the country. The AIDS Center is well known by its labs especially high performance, where both, screening (simple/rapid and ELISA methods) and confirmation methods (WB and PCR qualitative and quantitative) of HIV/AIDS and various other infections.

6. AIDS Center’s, Research Institute on Addiction's, databases of IDUs receiving VCT and records from other law threshold services; available in all selected cities except Telavi Georgian AIDS and Clinical Immunology Research Center provides VCT services to all interested individuals, including IDUs. The law threshold services of Research Institute on Addiction and local NGOs operate under the framework of different international projects. The program managers from all these services run the computer based database for monitoring of the program operation: # of first time service users, # of repeated users, information on risky behaviors of clients, utilization of commodities, etc.

Low threshold agencies often view drug misuse, and therefore the treatment of drug misuse as a social rather than a medical problem, and thus could be attracting a more

19 Article 273 of the Criminal Code of Georgia – “Illegal production, purchase, storage of narcotic drugs, their analogs or precursors for personal use and/or illegal use without doctor’s prescription”

representative group of drug misusers. These agencies may collate the same standard of information on their clients as the more formal drug treatment agencies described above, although in some instances some clients may only be known by a forename or an assumed name. The needle exchange programs provide basic supplies (syringes, needles, condoms, etc) to their clients on continuous basis. Along with the needle exchange the IDUs receive the information and counceling on safe injection and sexual practices. The VCT centers provide HIV risk reduction counseling to their clients. Relevant IEC materials and condoms are distributed as well by these services.

7. National Forensic Expertise Bureau of the Ministry of Justice database on IDUs deceased due to a fatal drug overdose In 2004, National Forensic Expertise Bureau was established at the Ministry of Justice, which re-started to register drug-related death cases. The data on fatal drug overdose in 2007 related only to cases investigated and tested by the Bureau in Tbilisi. Though the data do not cover the country in general and do not allow to be broken down according to the type of drug/s that caused the overdose, it is the first time when the Bureau broke the long drug death-related silence in Georgia. Data on whole Georgia are not yet available. Data Entry and Analysis The data was analyzed using Respondent Driven Sampling Analysis Tool version 6.0 (RDSAT). The sampling frame for RDS is based on specific information collected from participants, including: ⇒ Who recruited whom (tracked in RDSCM) ⇒ The relationship of the participant to the recruiter. The RDS population estimates are based on an assumption that the recruiter and the participant know each other. ⇒ The participant’s personal network size (i.e., how many injectors they know). The network size information from individuals is used to estimate the average network size by different sample characteristics (e.g., by gender, race/ethnicity, drug of choice, etc.).

From this frame, sampling probabilities can be calculated and, in turn, population estimates can be assessed for bias and the variability of these estimates can be determined. To calculate the population estimates derived from RDS, several sources of bias are taken into account: the differences in effective recruitment across groups (those more effective at recruitment would be overrepresented in the sample); homophily (groups that are more insular would be overrepresented because it is more difficult to break out of those groups); and the network size (groups with larger networks would be overrepresented because more recruitment paths lead to their members).

The researchers assessed whether the sample reached equilibrium, resulting, therefore, in a sample which should allow the calculation of unbiased population-based estimates. The parameters used to calculate the RDS population-based estimates were 15,000 bootstraps and imputation of 5 percent of the outliers in both extremes for the restricted network size. The number of recruitment waves required was calculated in RDSAT for all independent and key dependent variables. Almost all variables reached equilibrium between the third and fourth recruitment waves, and the remaining at the maximum of eleven waves. Raw data was first prepared using SPSS version 13.0. This included generating new variables, re-coding missing values following analysis strategy and RDSAT Manual. Datasets were then converted to Microsoft Excel files, and then to RDS files. RDS database was developed by Curatio International Foundation research team and kindly provided to us.

Socio-demographic variables in this study are presented as both sample proportions and population-based estimates with 95% confidence intervals (CI) weighted for personal network size and recruitment patterns based on RDSAT. Additionally, data for the multiplier calculation was also

35 analyzed using SPSS version 13.0 (95% confidence intervals (CI) were calculated by the SPSS Syntax Editor, Syntax for confidence intervals).

At the completion of the nomination interviewing process, database and statistic processing specialists created a database matching the questionnaire that included variable names, variable descriptions and value labels. The completed questionnaires were double entered, cleaned, processed and analyzed. Two experienced individuals made the data entry, one who read the completed interview form and the other entering the data. Once the SPSS databases were completed, a random check was made of 5% of the completed interview forms. In addition, a frequency was run on all variables to examine values, labels and frequencies. The “cleaned” database was submitted to Bemoni for data analysis. Quality Control The interviewing process was closely monitored by Project Coordinator and Expert Team Leader in all sites. The survey quality control was implemented through two stages: ⇒ Control of interviewing processes through site visits; ⇒ Attendance at interviews.

The findings of quality control show the surveys were undertaken in compliance with the existing instructions and no errors were reported. Specifically: ⇒ All respondents were interviewed in separate rooms; ⇒ All respondents were asked whether they had any objection to the interview and were explained the meaning of confidentiality; ⇒ All interviewers did their best to be polite and respectful; ⇒ All questions were asked in compliance with the written text; ⇒ Interviews were held at a pace set by the respondent.

FINDINGS

The findings of the study conducted in five locations across the country have bee presented in four sub-sections: 1. RDS Survey The discussion in this section will be centered on some key indicators such as age, gender, educational level, marital status, the use of drugs, and the exposure to different services. Major characteristics of study sample (total n =1127) and RDS-adjusted population-based estimates (95% CI) are presented in Tables 32-36 (see appendix 2).

⇒ Demographic and Social Characteristics of Respondents Gender and Age Virtually all (99,3% in Tbilisi, 97,7% in Gori, 100% in Telavi, 99,5% in Zugdidi and 98,1% in Batumi) IDUs interviewed were men. Only 15 women (3 in Tbilisi, 5 in Gori, 1 in Zugdidi and 6 in Batumi) were identified in the RDS methodology. Since there were few women recruited in these studies, the findings primarily represent male IDUs.

Majority of the respondents across the survey locations were in the age group of 31 to 40 years, except Tbilisi where the majority of respondents were between 41-50 years of age. The proportion of respondents in this age group was highest in Zugdidi (37.9%) and Batumi (37.5%) while it was lowest in Tbilisi (26.5%). Few IDUs are younger than 24 years of age in Tbilisi (7.6%) while this percentage is higher in other locations - about 23.1% in Gori, 18.3% in telavi, 15.4% in Zugdidi and 12.5% in Batumi. The proportion of respondents was the lowest in 50+ age group (Figure 5).

18-24 25-30 31-40 41-50 50+

40 35 30 25 20 15 10 5 0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 6. Distribution of Respondents by Age Groups

The median age of the respondents ranged between 32 and 40 years across the survey locations. The median age was observed to be highest in Tbilisi (40 years) while it was lowest in Telavi (32 years). In other cities (Gori, Zugdidi and Batumi) the median age was 34-35.

Educational Level Figure 6 presents the educational status of respondents across the survey locations, which includes the respondents with secondary, incomplete high and university education. In general, 37

IDUs tend to be well educated. Overall, the proportion of respondents with university digree was highest in Tbilisi (67.6%), followed by Zugdidi (36.6%), than comes Telavi and Gori ((30% and 28.9% respectively), and the lowest proportion was reported in Batumi (22.8 %). Among the whole survey population only 3 respondents (2 in Gori and 1 in Telavi) had not completed either a secondary schooling or vocational training.

Primary Secondary Incomplete high High

0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 7. Distribution of Respondents by Educational Level

Marital Status IDUs tend to be either single (never married) or married. Almost half of respondents reported being currently married. Virtually all of the married IDUs live with their spouse. 25% in Tbilisi, 36% in Gori, 48% in Telavi, 45% in Zugdidi and 39% in Batumi of study participants had never been married. Only few reported that they are divorced (Figure 7).

Married Divorced/separated Widower Never beeen married

0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 8. Distribution of Respondents by Marital Status 38

⇒ Drug Ussage Pattern All the IDUs interviewed across the five survey locations were asked their age when they first started taking any drug. They were also asked how long they had been injecting drugs, how old they were when they first took any injectable drug, frequency of drug injection in last week before the survey and type of drugs taken during last week.

Age at which started using drugs Figure 8 presents the findings on age at which the respondents started using any drug. Almost two- third (63% in Tbilisi, 68% in Gori and Telavi, 71% in Zugdidi and 59% in Batumi) of IDUs began using drugs between 15 to 19 years of age. About 25 percent of the respondents In Tbilisii and 28 percent in Batumi reported that they started using drugs before the agge of 15 years. In other cities the percentage of IDUs started using drugs before 15 years old varies between 12-14%. The median age of starting drugs was reported between 16-17 in all survey locations.

<15 15-19 20-24 25+ Mean

80 70

60 50

40 30

20 10

0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 9. Age When First Used Any Drug Age at which Started Injecting Drugs Almost half of all IDUs in all survey sites except Telavi began injecting drugs when they were between 15-19 years of age; In Telavi, about one out of three respondents (35%) began injecting drugs during between these ages. Figure 9 preseents the age distribution.

<15 15-19 20-24 25+ Mean

0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 10. Age When First Injected Any Drug

The median age of starting injecting drugs was also calculated for each survey location and it ranged from 18 years in Zgdidi to 20 years in Telavi and Gori.

Duration of Injecting Drugs Table 9 presents the average duration of drug injection. The mean varied from 12 years in Tbilisi to around 7 years in Gori.

Table 9. Mean duration of injecting drug use

RDS Population Estimates % (95% CI)

Indicators Tbilisi Gori Telavi Zugdidi Batumi

11.93 7.42 8.2 9.8 9.9 Mean duration of injecting (0.4-37) (0.2-39) (0.5-30) (0.5-30) (0.5-40) drug use (in years)

Membership of regular injecting group IDUs were also asked if they were a member of a regular injecting group in the last 6 months, and if so, how many members regularly injected together. Almost 3 of every 5 IDUs were the members of a regular injecting group. The size of these injecting groups ranged from a low of 2 to 20 members, for an average of from 3.8 (Batumi) to 4.67 (Tbilisi) members (Table10).

Table 10. Mean number of injecting group members

RDS Population Estimates % (95% CI)

Indicators Tbilisi Gori Telavi Zugdidi Batumi

4.67 4.62 4.5 4.6 3.8 Mean # of injecting group (2-15) (2-15) (2-10) (2-15) (2-10) members

Type of drugs injected last week Out of those who reported that had injected drugs last week, most of them stated that they inject several times a week. In addition to understanding the frequency of drug use, it is imperative to know the type of drugs used by the respondents. Thus, all the respondents were asked to recall all the type of drugs which they had injected in last week. Respondents could list more than one response in this multiple response question and the major findings are presented in figure 10.

CNS Depressants CNS Stimulants Narcotic drugs Other psychoactive substances Combination

100 90 80 70 60 50 40 30 20 10 0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 11. Injected Drugs Last Week

The highest percentages of the IDUs who had injected in the previous week injected Heroin, Buprenorphine (subutex), and Ephedrone. Among Narcotic Drugs Buprenorphine was reported mostly in Tbilisi (50.5%) and Gori (33%) while Heroin was widely used in Batumi (82%), and Telavi (36.4%). Illegal consumption of Morphine was observed in Gori (9.8%). Ephedrone (CNS Stimulants) was used more in Gori (38.3%), while antihistamines were reported by 39.6 percent of the respondents in Tbilisi (combination).

Of those IDUs that had injected drugs in the previous week they injected from 1.1 to 1.46 drugs, on average (1.46 in Tbilisi, 1.17% in Gori, 1.2% in Telavi and Zugdidi and 1.1% in Batumi).

⇒ Treatment/intervention history The exposure to drug treatment (detoxification, substitution) and different prevention services such as VCT, IEC, needle/syringe exchange was assessed in this survey.

Detoxification treatment for IDUs All the respondents were also asked whether they received any treatment for drug use or not and the current status of their treatment. There was considerable variation in reported cases of exposure to detoxification treatment across the study locations. Maximum proportion of respondents reported that they have never received any treatment. This proportion was reported to be highest in Gori (84.8%) and lowest in Batumi (51.6%). Consequently, the proportion of respondents who underwent detoxification treatment was reported highest in Batumi (43.9%) and and lowest in Gori (1.2%). This proportion was 39.4% in Tbilisi, 29.5% in Telavi and 37.5% in Zugdidi.

Currently taking medical treatment Used to take medical treatment, but quit Never have been treated

90 80 70 60 50 40 30 20 10 0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 12. Detoxification Treatment

HIV/AIDS Program intervention in the last 12 months The proportion of respondents who attended methadone substitution program was 1.9% in Tbilisi and 1.4% in Batumi. More than a half of the respondents were given information about this program. The proportion of IDUs who have heard information about needle/syringe exchange varies from 15% in Tbilisi to 50% in Gori. 18.3% of respondents in Gori reported that they were given sterile syringes in last year. The lowest percentage of IDUs by this indicator was in Zugdidi – 1.1%. The proportion of respondents who were given condoms in the last 12 months was reported to be highest in Batumi (25.6%) and Gori (20.3%) and lowest in Telavi (6%) and Tbilisi (9%).

The proportion of respondents who reported that it was possible for them to get a confidential HIV test was highest in Tbilisi (92.6%), and in Batumi (87.4%) and lowest in Telavi (65.8%). Among all locations proportion of respondents who had taken a voluntary HIV test and received the results in the last 12 months was highest in Gori (8.4%) and lowest in Telavi (2.9%). Figure 12 presents findings on exposure to various services in the last one year. Tested on HIV and received results Given condoms

Given IEC materials Given sterile needles/syringes

Used substitution therapy

35 30 25 20 15 10 5 0 Tbilisi Gori Telavi Zugdidi Batumi

Figure 13. Other Services

⇒ Biomarker The biomarker component of the survey involved the analysis of blood specimens at the Laboratory of Serology and Virology of the AIDS Center in Tbilisi.

HIV testing: HIV antibody testing was performed using a three-level enzyme-linked immunosorbent assay (ELISA) testing strategy. If a sample was reactive in the first ELISA (Genescreen Plus HIV Ag-AB, Bio-rad) test, the sample was retested two more times using another kit of ELISA. Samples were considered HIV antibody positive if they were reactive in two out of three tests. Any sample non-reactive to the first test was considered as HIV-antibody negative. HIV-antibody positive samples were tested with Western Blot (HIV blot, Genelabs) as the confirmatory test for HIV. Syphilis testing: Serum samples were tested also for syphilis antibodies with rapid plasma regain (RPR, Human) test and Treponema pallidum hemaglitination assay (TPHA, Human). ELISA (ELISA TP IgG test [Nubenco]) tests were used for confirmation of syphilis-antibody positive samples.

Table 11. Prevalence of HIV and STIs

RDS Population Estimates % (95% CI)

Indicators Tbilisi Gori Telavi Zugdidi Batumi

2.5 0 1.5 2.2 4.5 HIV (0.3 – 5.4) (0-3.5) (0-3.5) (1.5-8)

6.3 3.9 5.5 6.9 7.6 Syphilis (3.7 – 9.3) (1.1 – 7.3) (2.5-8.5) (3.5-11) (4-12)

2. Benchmark Data As a drug user may be in contact with more than one agency, and therefore be included in the data from more than one source, sufficient information is needed on each individual to identify multiple occurrences. Matching records between data sources can be complex, and within the area of record linkage, it is recognized that problems exist even when several different fields of data on each individual has been collected.

Health-related Indicators

⇒ Injection drug users (IDUs) in abstinence oriented treatment in 2007

Source of information: National Center for Disease Control and Public Health (NCDC)

Table 12. Detoxification treatment benchmark data

City Treatment Facility # of # of Total # Inpatient Outpatient IDUs IDUs Tbilisi Research Institute on Addiction 229 285 514 Center “Uranti” 203 75 278 Clinic “Bemoni” 119 10 129 Addiction Dispensary 154 26 180 Clinic “Tanadgoma” 82 15 97 Total in Tbilisi 787 411 1198 43

Batumi Addiction Center 47 14 61 Center for Medical Support and 0 57 57 Expertise “Levgori” Total in Batumi 47 71 118 Grand Total 834 482 1316

Explanation: Double counting cannot be excluded, as many drug users will come into contact with a variety of treatment facilities. Utilizing unique personal identifiers to prevent double counting is impossible in Georgia.

The absolute majority (96%) of the individuals treated (detoxified) in 2007 were opioid users. Out of 834 IDUs treated in 2007, 769 were diagnosed as opioid dependents (F 11 – Mental and behavioral disorders due to use of opioids, ICD-10), in 32 cases multiple drug use (F 19 – Mental and behavioral disorders due to multiple drug use: opioids + other psychoactive substances) was diagnosed and 33 patients received treatment due to injecting of Ephedrone (F 15 – Mental and behavioral disorders due to use of other stimulants). All outpatients were opioid dependents.

⇒ Injection drug users on waiting lists for the abstinence oriented treatment in 2007

Source of information: Drug treatment agencies

Explanation: As long as the capacity of facilities was limited, drug users had been put on waiting lists. They might break off contact again, before any data can be collated, or they may leave incomplete, unreliable or even wrong data. There is no official data on this. Based on personal communication with the heads of clinics it can be assumed that in 2007 the average number of IDUs on waiting lists in the abstinence oriented treatment facilities in Tbilisi was 1200. There were no waiting lists in Batumi.

⇒ Drug users in Methadone substitution treatment in 2007

Source of information: Methadone Substitution Programme database of the Georgian Research Institute on Addiction

Table 13. Methadone substitution treatment benchmark data

City Treatment Facility # of # of Total # Among Male Female them, # of IDUs IDUs HIV+ IDUs Tbilisi Research Institute on 94 0 94 14 Addiction Center “Uranti” 91 1 92 13 Total in Tbilisi 185 1 186 27 Batumi Addiction Center 101 0 101 7 Total in Batumi 101 0 101 7 Grand Total 286 1 287 34

Explanation: Three pilot Methadone substitution therapy programs were operated in Georgia in 2007. These programs had been coordinated by the Georgian Research Institute on Addiction.

⇒ Injection drug users on waiting lists for Methadone substitution treatment in 2007

Source of information: Methadone Substitution Programme database of the Georgian Research Institute on Addiction

Table 14. Methadone substitution waiting list benchmark data

City # of IDUs at # of IDUs # of IDUs on waiting Out of them, # of the during the year lists in 2007 IDUs included into the beginning 2007 Program during the of year 2007 year 2007 Male Female Tbilisi 125 160 282 3 55 Batumi 22 154 176 0 101 Grand Total 147 314 461 156

⇒ Drug users using needle exchange and other low-threshold programs in 2007

Source of information: Monitoring systems of low threshold agencies - computer based database for monitoring of the program operation

Table 15. Needle exchange benchmark data

Needle Exchange City Needle # of “Permanent” # of Other Clients Total # of Exchange Clients Clients Programs Male Female Male Female Tbilisi NGO “New 451 28 261 51 791 Wave” NGO “New 171 4 895 72 1142 Vector” Total in Tbilisi 622 32 1156 123 1933 Batumi NGO “Imedi” 483 7 2224 0 2714 Gori NGO “Nabiji 295 15 330 20 660 Momavlisken” Zugdidi NGO “ORDU” 170 0 96 0 266 Grand Total 1570 54 3806 143 5573

Table 16. VCT services benchmark data

VCT City VCT Centers # of Clients # of Clients reached Total # of received VCT in by outreach workers Clients Centers Male Female Male Female Tbilisi AIDS Center 365 15 380 Research Institute on 927 32 1468 0 2427 Addiction NGO “New Wave” 655 624 3923 0 5202 NGO “Bemoni” 452 8 460 NGO “Tanadgoma” – 412 0 412

in prisons Total in Tbilisi 2399 679 5803 0 8881 Batumi NGO “Tanadgoma” 128 0 128 Addiction Dispensary 224 0 224 Total in Batumi 352 0 352 Zugdidi NGO “Ksenoni” 808 4 812 Grand Total 3559 683 5803 0 10,045

Explanation: The main services offered to IDUs under the harm reduction programs in Georgia are voluntary counseling and testing (VCT), and needle exchange. These activities are carried out within three major projects: 1) Harm reduction programme of the “Open Society – Georgia” Foundation; 2) Anti drug component of the GFATM Programme “Strengthening national response in Georgia to implement the effective prevention and control of HIV/AIDS, tuberculosis and malaria”, and 3) STI/HIV Prevention (SHIP) Project for Georgia, funded by USAID. All the agencies providing these services to IDUs in 2007 are included in present study.

⇒ Drug users tested on HIV in 2007

Source of information: HIV/AIDS register run by the Georgian AIDS and Clinical Immunology Research Center

Table 17. HIV testing benchmark data

City Service Provider # of IDUs # of IDUs # of IDUs tested on infected by deceased HIV HIV due to AIDS Tbilisi AIDS Center 380 43 7 Research Institute on Addiction 460 5 Methadone substitution (2 200 36 Programmes) Center “Uranti” 108 2 NGO “New Way” 392 4 Center Tanadgoma (in prisons) 412 3 Total in Tbilisi 1952 93 7 Batumi Methadone substitution 101 7 2 Program Union “Imedi” 147 3 Healthy Cabinet 66 3 Outpatients of other clinics 23 0 Inpatients of other clinics 47 2 Total in Batumi 384 15 2 Zugdidi NGO “Ksenoni” 427 8 1 Gori Blood Transfusion Station 127 3 3 Telavi AIDS Center 25 1 0 Other cities 81 26 Grand Total 2915 201 39

Explanation: The cases are identified through routine surveillance data reported by HIV diagnostic labs operating throughout the country.

HIV antibody testing was performed using a three-level enzyme-linked immunosorbent assay (ELISA) testing strategy. If a sample was reactive in the first ELISA (Genescreen Plus HIV Ag-AB, 46

Bio-rad) test, the sample was retested two more times using another kit of ELISA. Samples were considered HIV antibody positive if they were reactive in two out of three tests. Any sample non- reactive to the first test was considered as HIV-antibody negative. HIV-antibody positive samples were tested with Western Blot (HIV blot, Genelabs) as the confirmatory test for HIV.

⇒ Drug users deceased due to a fatal drug overdose in 2007

Sources of information: National Center for Disease Control and Public Health (NCDC); National Forensic Expertise Bureau of the Ministry of Justice

Explanation: NCDC reported that the death cases caused by fatal drug overdose were not registered by the relevant agencies (emergency departments, ambulances) in the country in 2007. Although everybody knows that such cases exist, emergency rooms make other diagnoses and do not record the cases of drug use.

According to the National Forensic Expertise Bureau of the Ministry of Justice, 39 cases of drug overdose deaths had been recorded in Tbilisi, i.e. approximately 1.06% of all unnatural deaths in Georgia in 2007.

Crime-related Indicators

⇒ Injection drug users registered by the police tested positively for presence of illegal drugs in 2007

Source of information: Ministry of Internal Affairs

Table 18. Benchmark data on IDUs came into contact with the police

City Total # of registered drug of those, # of registered users, based on the positive IDUs, based on the positive test results test results Tbilisi 9519 8168 Telavi 82 28 Zugdidi 1166 381 Gori 1602 829 Batumi 2634 2244 Grand Total 15,003 11,650

Explanation: Taking into consideration that Georgian drug legislation does not distinguish between being detained in connection with the use of drugs and being convicted for purchase or possession of drugs, we use only police records regarding the persons tested positively for presence of illegal drugs.

3. Calculation of the Size of IDU Population in 5 selected cities Multipliers were derived from the RDS survey of 1127 IDUs recruited from across 5 cities. Participants’ responses to the questionnaire were used to produce a final series of IDU size estimates, including 95% confidence intervals.

The following section provides specific estimates for each selected city. Different number of separate multiplier estimates was made to calculate the quantity of problem drug users in different cities.

Calculation of the estimated size of the IDU population in the surveyed cities revealed these figures (mean estimates): Tbilisi - 27 107 (23 694-31 532); Gori – 2 989 (2 537-3 570); Telavi – 557 (358- 941); Zugdidi - 4 855 (3 945-6 089); Batumi – 5 937 (5 008-7 162). Tables 19-23 set out the multiplier estimates of IDUs in 5 cities across the country derived from different sources, together with the mean and median of the estimates.

Table 19. Estimates of the number of IDUs in Tbilisi

Estimated Benchmark Multiplier 95% CI size 95% CI Police data 8168 2.54 2.37 2.74 20747 19358 22380 HIV testing data 1952 7.06 6.13 8.18 13781 11966 15967 Treatment data 2398 3.6 3.28 3.96 8633 7865 9496 Methadone substitution data 471 4.64 4.17 5.21 2185 1964 2454 Low Threshold Programs data 10814 8.34 7.15 9.82 90189 77320 106194 Mean 27107 23694 31532 Median 13781 11966 15967

Table 20. Estimates of the number of IDUs in Gori

Estimated Benchmark Multiplier 95% CI size 95% CI Police data 829 4.27 3.73 4.92 3540 3092 4079 HIV testing data 127 11.66 9.17 15.11 1480 1165 1919 Low Threshold Programs data 660 5.98 5.08 7.14 3947 3353 4712 Mean 2989 2537 3570 Median 3540 3092 4079

Table 21. Estimates of the number of IDUs in Telavi

Estimated Benchmark Multiplier 95% CI size 95% CI Police data 28 4.09 3.55 4.76 114.52 99.4 133.28 HIV testing data 25 40 24.63 69.93 1000 615.75 1748.25 Mean 557 358 941 Median 557 358 941

Table 22. Estimates of the number of IDUs in Zugdidi

Estimated Benchmark Multiplier 95% CI size 95% CI Police data 381 3.16 2.82 3.57 1204 1074 1360 HIV testing data 427 7.08 5.86 8.69 3023 2502 3710.63 Low Threshold Programs data 1078 9.59 7.66 12.24 10338 8258 13195 Mean 4855 3945 6089 Median 3023 2502 3711

Table 23. Estimates of the number of IDUs in Batumi

Estimated Benchmark Multiplier 95% CI size 95% CI Police data 2244 2.03 1.88 2.22 4555 4219 4982 HIV testing data 384 5.84 4.89 7.06 2243 1878 2711 Treatment data 118 4.28 3.7 5.01 505 437 591

Methadone substitution data 277 6.64 5.49 8.16 1839 1521 2260 Low Threshold Programs data 3066 6.7 5.54 8.24 20542 16986 25264 Mean 5937 5008 7162 Median 2243 1878 2711

4. Estimation of the prevalence of injection drug use

Prevalence estimates for the injection drug use were produced for 5 cities of Georgia. Census data gave the population between 18 and 64 for urban areas across the country. The appropriate estimations of injecting drug use shown in the tables above were then applied to that population. The statistical lower and upper limits (at 95% confidence interval) were used to reflect the minimum and maximum ranges.

Tables 24-28 below present the IDU prevalence estimation (%) in 5 cities across the country derived from different sources, together with the mean and median of the estimates.

Table 24. Estimated Prevalence Rates in Tbilisi

Tbilisi Adult population (18-64) 672000 Estimated Prevalence of IDU size 95% CI (%) 95% CI Police data 20747 19358 22380 3,09 3,04 3,13 HIV testing data 13781 11966 15967 2.05 2.01 2.09 Treatment data 8633 7865 9496 1.28 1.25 1.32 Methadone substitution data 2185 1964 2454 0.33 0.31 0.34 Low Threshold Programs data 90189 77320 106194 13.42 13.33 13.51 Mean 27107 23694 31532 4,03 3,98 4,09 Median 13781 11966 15967 2,05 2,01 2,09

Table 25. Estimated Prevalence Rates in Gori

Gori Adult population (18-64) 82800 Estimated Prevalence of IDU size 95% CI (%) 95% CI Police data 3540 3092 4079 4,28 4,12 4,43 HIV testing data 1481 1165 1919 1.79 1.69 1.89 Low Threshold Programs data 3947 3353 4712 4.77 4.6 4.93 Mean 2989 2537 3570 3,61 3,47 3,75 Median 3540 3092 4079 4,28 4,12 4,43

Table 26. Estimated Prevalence Rates in Telavi

Telavi Adult population (18-64) 42900 Estimated Prevalence of IDU size 95% CI (%) 95% CI Police data 114.52 99.4 133.28 0.27 0.21 0.33 HIV testing data 1000 615.75 1748.25 2.33 2.17 2.5 Mean 557 358 941 1,30 1,19 1,42 Median 557 358 941 1,30 1,19 1,42

Table 27. Estimated Prevalence Rates in Zugdidi

Zugdidi Adult population (18-64) 105000 Estimated Prevalence of IDU size 95% CI (%) 95% CI Police data 1204 1074 1360 1,15 1,07 1,22 HIV testing data 3023 2502 3711 2.88 2.77 3 Low Threshold Programs data 10338 8258 13195 9.85 9.64 10.05 Mean 4855 3945 6089 4,63 4,37 4,76 Median 3023 2502 3711 2,88 2,77 3,00

Table 28. Estimated Prevalence Rates in Batumi

Batumi Adult population (18-64) 74500 Estimated Prevalence of IDU size 95% CI (%) 95% CI Police data 4555 4219 4982 6,11 5,92 6,31 HIV testing data 2243 1878 2711 3.01 2.87 3.15 Treatment data 505 437 591 0.68 0.61 0.75

Methadone substitution data 1839 1521 2260 2.47 2.34 2.6 Low Threshold Programs data 20542 16986 25264 27.57 27.21 27.94 Mean 5937 5008 7162 7,97 7,79 8,15 Median 2243 1878 2711 3,01 2,87 3,15

EXTRAPOLATION FROM LOCAL TO NATIONAL PREVALENCE ESTIMATES

Local estimates using multiplier-benchmark methods give important information on extent of drug problem. However, they are employed in studies of drug use on a smaller, geographically local scale. Nonetheless, there is still very often a need for overall national estimates to be made, and one way of doing that is to extrapolate from local prevalence studies to an overall picture.

Extrapolation methods are not a specific method of prevalence estimation in themselves, but when some prevalence information is known they are used to extend that information into areas - usually, other geographic regions—where the prevalence information is not known. The important element of any extrapolation method is that it makes use of known prevalence figures in certain regions to estimate prevalence in other regions. To do that, the regions must have some data sources that are the same as (or very similar to) the regions for which prevalence estimates exist, although of course they lack the regional prevalence figure itself. The general principle is then to use data that are similar across the separate localities to project figures for drug use prevalence from localities where it is known to localities where it is lacking.20

The extrapolation methods are based on statistical regression techniques. The method described below comes under various headings: usually, “synthetic estimation”, or “multi-indicator” method, or sometimes under the more technical name of “regression on principal components“.

Extrapolation Method: The multivariate indicator for injection drug use21 The Multivariate Indicator Method (MIM) is a special case of synthetic estimation. Generally, synthetic estimation methods are methods which transfer information about a variable of interest, e.g. drug use prevalence, from a population in which it can be observed (calibration population/anchor point) to a target population in which it cannot be observed. From anchor points, a functional relationship between some variables and the variable of interest is derived which is extended to the target population. Applied to the field of drugs, the prevalence of problem drug use in a country may be estimated by relating a set of drug use indicators, which are available in all regions of a country, to prevalence estimates in a few regions (calibration population). The indicators may be directly (e.g. mortality, morbidity, and arrest) or indirectly related to drug use (e.g. population density, unemployment rate, housing density). Typically, analyses are based on prevalence rates and indicator rates per 100,000 inhabitants.

With regard to the MIM, two main variants of the method are common. One way is to estimate the relationship between drug use indicators and prevalence estimates in the anchor points via (linear) regression and to apply the regression coefficients to the drug use indicators in the target population. This yields prevalence estimates for the non-anchor points. Summing up all regional prevalence estimates yields the national prevalence estimate. Smit and colleagues (2003) used this method to estimate local and national

20 Estimating Prevalence: Indirect Methods for Estimating the Size of the Drug Problem. Global Assessment Programme on Drug Abuse, Toolkit Module 2. UNODC, 2003

21 Key Epidemiological Indicator: Prevalence of problem drug use, EMCDDA, 2004

52 problem drug use prevalence in the Netherlands, employing population density and housing density as indicators.

As the anchor points have a great impact on the actual figures of the total prevalence by fixing the regression line, great care has to be taken in obtaining reliable and valid estimates with the same target group. Furthermore, the estimates should cover at least one area with an assumed high prevalence rate and at least one region at the lower end of prevalence rates, in order to improve the quality of the regression model. Using only estimates of regions with a high prevalence makes the method useless, and may even result in negative prevalence rates. Indicator values for the anchor points must be available. In practice, prevalence estimates are often available only on city level whereas indicators are collected on a regional level. If problem drug use is concentrated heavily in these cities they may be used as anchor points. However, the relationship between indicators and drug use prevalence may be different for urban and rural areas. Application The aim of this method is to estimate the number of problem drug users in the population by combining information on prevalence that is available only in a few areas (the calibration population, or anchor points) and indicators or predictors of drug use that are available in all areas (Mariani and others (1994)). The method was first used in the United States (Woodward and others (1984)) and has been described more fully elsewhere (Wickens (1993)).

The key assumption of the method is that the relationship between prevalence (dependent variable) and the predictors (independent variables) in the calibration sample is transferable to all other areas. It is also assumed that a single factor underlies the drug- related indicators and that principal components analysis can be used to extract the main factor that explains the largest amount of variance in the indicators.

The application of the multivariate indicator method requires a breakdown of national states by regions or provinces and data on problem/injection drug use (indicators), which must be available for each of the regions and refer to the same time period. The national IDU prevalence estimates in the present study were derived from the estimates of the urban areas. Since injection drugs are more available in cities and drug injection is not common in rural areas (locally cultivated pot is particularly widely spread in villages), actually there is a little number of IDUs in rural areas as well. Consequently, not considering this population may have resulted in an under-estimate. However, assuming that injection drug users are mainly concentrated in the urban parts of Georgia we are willing to ignore this downward bias.

Two separate national estimations were produced: Estimation N 1. It is recommended to use drug-related indicators as predictors in this regression model, i.e. drug related offences, drug-related deaths, clients in treatment, HIV cases related to injection drug use, imprisoned drug users (EMCDDA, 1999). Unfortunately, however, these statistics are not available in Georgia for the whole country. Due to a lack of available drug-related indicators the Dutch research group used an alternative model with social indicators such as housing density and population

density.22 Similarly, taking in consideration that none of the drug-related indicators could be obtained for all urban areas in Georgia, national IDU prevalence was calculated using only one demographic indicator such as population density (Census data). Unfortunately the data on housing density was not collected in the Country.

Estimation N 2. The second method used the drug injection prevalence rate coefficient for each city in order to estimate the number of injection drug users nationwide (modified from the method suggested by E. Pizani). 23 It was based on input from people working in the area of drug addiction. Addiction experts ranked all 65 cities in Georgia by prevalence rates with corresponding coefficients. Five categories of prevalence rate coefficients had been chosen and each city was assigned to one of the following categories:

Prevalence Very High High Medium Low Very Low Rate Coefficient 8 3,5 1,5 1,0 0,5

Description of the Multivariative Indicator Method Applied Five indicators, denoted by A, B, C, D and E had been used for MIM. Additionally to the indicators, the population size F of the age group 18-64 in each city (totally 65 cities) as well as independently obtained prevalence estimates G for 5 cities (the so-called anchor points) are needed.

The different indicators highlight different aspects of the drug problem. No indicator is supposed to measure prevalence. The indicators are, however, indicative of whether problem drug use increases or decreases (Person et al., 1977). By applying principal component analysis a common factor is extracted which is assumed to be proportional to prevalence of problem drug use. As principal component analyses underlies the assumption of a linear relationship between observable variables and the principal components there should be a linear relationship between indicators of problem drug use and the unknown prevalence.

Obviously, the validity of prevalence estimation can be improved by increasing the number of anchor points. Then, more drug use indicators (proxy variables) can be used in the linear regression model. One of the problems is, however, the choice of appropriate drug use indicators (proxy variables). If the number of drug use indicators equals or exceeds the number of anchor points linear regression is not possible. As drug use indicators are more easily available than reliable regional prevalence estimates it is often necessary to reduce the number of drug use indicators. Up to now, different methods of reducing the number of indicators have emerged: Mariani (1999) as well as Person, Retka and Woodward (1977, 1978) applied a principal component analysis (PCA).24

22 Estimating Local and National Problem Drug Use prevalence from Demographics, Filip Smit et al., Addiction Research and Theory, 2003, Vol. 11, N6

23 Estimating the number of drug injectors in Indonesia. Elizabeth Pisani, International Journal of Drug Policy N 17, 2006

24 Prevalence of problem drug use at the national level, EMCDDA, 2002

The steps below summarize the process used to derive the national estimate for the percentage of injection drug users in Georgia using the Multiple Indicator Method.

Step 1. Data indicating the prevalence of injection drug use must be collected for a defined time period for each city. The following variables were used as indicators: A - Number of IDUs registered by Police for drug consumption B - Number of IDUs tested on HIV C - Number of clients in treatment D - Number of clients of the law threshold services E - Population density (for the estimation N 1) and prevalence rate coefficients (for the estimation N 2).

Step 2. In addition, the population size F for urban areas had been obtained from data on population distribution (State Department of Statistics of the Ministry of Economic Development of Georgia).

Step 3. For five selected cities reliable independent estimates G (resulting from the multipliyer- benchmark study) are necessary. These cities are called „anchor points“.

Step 4. For each of the variables A to E, G and for each region the figure per 100,000 inhabitants has to be calculated. AF=A*100,000/F GF=G*100,000/F

Step 5. Principal components analysis requires standardised values for AF to GF (subtracting the mean and dividing by the standard deviate).

Step 6. Principal components analysis of AF to EF with the extraction of the first factor, whose coefficients are saved. No rotational solution is needed, as any rotation only serves as an improvement for the fit of a set of indicators, and is therefore here redundant as only one indicator will be extracted.

Step 7. A linear regression (dependent variable: GF, independent variable: coefficients of the first factor) results in estimated prevalence rates per 100,000 inhabitants. Finally, these have to be transformed to prevalence estimates for the cities (multiplying with F and dividing by 100,000). Summation of the urban area prevalence estimates yields the national prevalence estimate.

In order to derive national estimates original data was entered into the SPSS version 13.0 data files, than SPSS-Syntax of the variant "PCA per 100,000" reflecting the above mentioned steps had been created based on instructions provided in the EMCDDA Scientific Report.25 The regression analysis was done by this SPSS syntax to make predictions of the estimated level of the drug abuse prevalence rates. Two separate estimations (by demographic indicator and by prevalence rate coefficients) were made.

25 Prevalence of problem drug use at the national level, EMCDDA, 2002

Results of the national prevalence estimation National prevalence estimates for the injection drug use were produced for 65 cities of Georgia. Census data gave the population between 18 and 64 for all urban areas across the country. Calculation of the IDU prevalence estimation nationwide revealed these figures: estimation method N 1, using demographic indicator (population density) – 1,46% (estimated number of IDUs equals 39 152) estimation method N 2, using prevalence rate coefficients - 1,53% (Number of IDUs – 41 062).

Tables 29-30 below present the national IDU prevalence estimation (%) produced by 2 different indicators: Table 29. Estimation by Population Density

Density of the Total Population Prevalence Prevalence Estimated N Cities Population per Population 18-64 per 100 000 % Number 1 sq.km

1 Tbilisi 1 081 679 659 824 4425.8 3 856,2 3,86 25 913,7 2 Batumi 121 806 74 302 7293.8 8 910,4 8,91 6 638,2 3 Keda 20 024 12 215 44.3 88,1 0,09 10,8 4 Kobuleti 88 063 53 718 122.3 62,9 0,06 33,8 5 Shuakhevi 21 850 13 329 37.2 72,5 0,07 9,7 6 Khelvachauri 90 843 55 414 219.8 94,4 0,09 52,3 7 Khulo 33 430 20 392 47.1 63,5 0,06 12,9 8 Lanchkhuti 40 507 24 709 76.0 77,8 0,08 19,2 9 Ozurgeti 78 760 48 044 144.4 201,0 0,20 96,55 10 Chokhatauri 24 090 14 695 29.2 57,5 0,06 8,4 11 Kutaisi 185 965 113 439 2746.9 722,2 0,72 819,27 12 Baghdati 29 235 17 833 35.9 57,9 0,06 10,3 13 Vani 34 464 21 023 61.9 75,3 0,08 15,8 14 Zestaponi 76 208 46 487 180.2 92,7 0,09 43,1 15 Terjola 45 496 27 753 127.4 106,0 0,11 29,4 16 Samtredia 60 456 36 878 166.0 104,4 0,10 38,5 17 Sachkhere 46 846 28 576 48.1 51,8 0,05 14,8 18 Tkibuli 31 132 18 991 65.0 84,2 0,08 16,0 19 Tskhaltubo 73 889 45 072 116.9 68,8 0,07 31,0 20 Chiatura 56 341 34 368 184.0 120,3 0,12 41,3 21 Kharagauli 27 885 17 010 30.5 53,8 0,05 9,2 22 Khoni 31 749 19 367 74.1 91,8 0,09 17,8 23 Akhmeta 41 641 25 401 18.9 34,3 0,03 8,7 24 Gurjaani 72 618 44 297 85.8 56,5 0,06 25,0 25 Dedoplis Tskaro 30 811 18 795 12.2 32,5 0,03 6,1 26 Telavi 70 589 43 059 84.4 263,6 0,26 113,1 27 Lagodekhi 51 066 31 150 57.4 54,8 0,05 17,1 28 Sagarejo 59 212 36 119 39.7 329,1 0,33 118,86 29 Sighnaghi 43 587 26 588 34.8 44,8 0,04 11,9 30 Kvareli 37 658 22 971 37.7 51,0 0,05 11,7 31 Akhalgori 7 703 4 699 7.6 50,6 0,05 2,4 32 Dusheti 33 636 20 518 11.3 30,6 0,03 6,3 33 Tianeti 14 014 8 549 15.5 54,2 0,05 4,6

34 Mtskheta 64 829 39 546 90.1 62,9 0,06 24,9 35 Kazbegi 5 261 3 209 4,9 48,9 0,05 1,6 36 Ambrolauri 16 079 9 808 16.3 51,4 0,05 5,0 37 Lentekhi 8 991 5 485 6.7 43,2 0,04 2,4 38 Oni 9 277 5 659 5.4 38,2 0,04 2,2 39 Tsageri 16 622,0 10 139 22.0 60,9 0,06 6,2 40 Poti 47 149 28 761 716.6 1 480,5 1,48 425,8 41 Abasha 28 707 17 511 89.0 115,2 0,12 20,2 42 Zugdidi 167 760 102 334 346.9 1 965,7 1,97 2 064,0 43 Martvili 44 627 27 222 50.7 55,1 0,06 15,0 44 Mestia 14 248 8 691 4.7 30,5 0,03 2,7 45 Senaki 52 112 31 788 100.1 79,1 0,08 25,2 46 Chkhorotsku 30 124 18 376 48.6 69,7 0,07 12,8 47 Tsalenjikha 40 133 24 481 62.1 67,7 0,07 16,6 48 Khobi 41 240 25 156 62.6 66,8 0,07 16,8 49 Adigeni 20 752 12 659 25.9 58,6 0,06 7,4 50 Aspindza 13 010 7 936 15.8 57,5 0,06 4,6 51 Akhalqalaqi 60 975 37 195 49.4 45,2 0,05 16,8 52 Akhaltsikhe 46 134 28 142 63.9 228,4 0,23 64,28 53 Borjomi 32 422 19 777 27.2 75,5 0,08 14,94 54 Ninotsminda 34 305 20 926 25.3 42,9 0,04 9,0 55 Rustavi 116 384 70 994 1920.5 980,3 0,98 695,96 56 Bolnisi 74 301 45 324 92.4 162,4 0,16 73,61 57 Gardabani 114 348 69 752 87.7 43,8 0,04 30,6 58 Dmanisi 28 934 17 650 23.4 45,1 0,05 8,0 59 Tetri Tskaro 25 354 15 466 21.6 46,4 0,05 7,2 60 Marneuli 118 221 72 115 126.4 53,1 0,05 38,3 61 Tsalka 20 888 12 742 19.8 49,4 0,05 6,3 62 Gori 148 686 90 698 146.7 1 528,0 1,53 1 265,2 63 Kaspi 52 217 31 852 65.0 58,5 0,06 18,6 64 Kareli 50 422 30 757 46.2 48,4 0,05 14,9 65 Khashuri 62 714 38 256 107.1 72,6 0,07 27,8 39 152,4

Table 30. Estimation by Prevalence Rate Coefficient

Total Population Prevalence Prevalence Prevalence Estimated N Cities Rank Population 18-64 Coefficient per 100 000 % Number

1 Tbilisi 1 081 679 659 824 H 3,5 4274,72476 4,28 28726,15 2 Batumi 121 806 74 302 VH 8,0 8531,66778 8,53 6356,09 3 Keda 20 024 12 215 VL 0,5 67,93795 0,68 8,3 4 Kobuleti 88 063 53 718 L 1,0 31,41563 0,31 16,88 5 Shuakhevi 21 850 13 329 VL 0,5 52,70972 0,53 7,03 6 Khelvachauri 90 843 55 414 VL 0,5 74,11766 0,74 41,07 7 Khulo 33 430 20 392 VL 0,5 5,13605 0,05 1,05 8 Lanchkhuti 40 507 24 709 VL 0,5 24,20531 0,24 5,98 57

9 Ozurgeti 78 760 48 044 L 1,0 163,03544 1,63 78,33 10 Chokhatauri 24 090 14 695 VL 0,5 37,1819 0,37 5,46 11 Kutaisi 185 965 113 439 M 1,5 248,72324 2,49 282,15 12 Baghdati 29 235 17 833 VL 0,5 10,52581 0,11 1,88 13 Vani 34 464 21 023 VL 0,5 8,41074 0,08 1,77 14 Zestaponi 76 208 46 487 VL 0,5 66,40416 0,66 30,87 15 Terjola 45 496 27 753 VL 0,5 34,08313 0,34 9,46 16 Samtredia 60 456 36 878 VL 0,5 53,92899 0,54 19,89 17 Sachkhere 46 846 28 576 VL 0,5 36,39434 0,36 10,4 18 Tkibuli 31 132 18 991 VL 0,5 2,92065 0,03 0,55 19 Tskhaltubo 73 889 45 072 VL 0,5 64,90146 0,65 29,25 20 Chiatura 56 341 34 368 VL 0,5 49,52083 0,50 17,02 21 Kharagauli 27 885 17 010 VL 0,5 16,56823 0,17 2,82 22 Khoni 31 749 19 367 VL 0,5 0,64294 0,01 0,12 23 Akhmeta 41 641 25 401 VL 0,5 26,65839 0,27 6,77 24 Gurjaani 72 618 44 297 VL 0,5 64,03715 0,64 28,37 Dedoplis 25 Tskaro 30 811 18 795 VL 0,5 4,14173 0,04 0,78 26 Telavi 70 589 43 059 M 1,5 287,65591 2,88 123,4 27 Lagodekhi 51 066 31 150 VL 0,5 42,83097 0,43 13,34 28 Sagarejo 59 212 36 119 L 1,0 389,34193 3,89 140,63 29 Sighnaghi 43 587 26 588 VL 0,5 30,57054 0,31 8,13 30 Kvareli 37 658 22 971 VL 0,5 17,39046 0,17 3,99 31 Akhalgori 7 703 4 699 VL 0,5 359,40278 3,59 16,89 32 Dusheti 33 636 20 518 VL 0,5 5,80452 0,06 1,19 33 Tianeti 14 014 8 549 VL 0,5 146,08498 1,46 12,49 34 Mtskheta 64 829 39 546 L 1,0 1,71661 0,02 0,68 35 Kazbegi 5 261 3 209 VL 0,5 579,27399 5,79 18,59 36 Ambrolauri 16 079 9 808 VL 0,5 112,64624 1,13 11,05 37 Lentekhi 8 991 5 485 VL 0,5 291,54511 2,92 15,99 38 Oni 9 277 5 659 VL 0,5 279,03384 2,79 15,79 39 Tsageri 16 622,0 10 139 VL 0,5 105,233 1,05 10,67 40 Poti 47 149 28 761 M 1,5 1300,35054 13,00 373,99 41 Abasha 28 707 17 511 VL 0,5 12,8214 0,13 2,25 42 Zugdidi 167 760 102 334 H 3,5 2313,26483 2,31 2428,93 43 Martvili 44 627 27 222 VL 0,5 32,52142 0,33 8,85 44 Mestia 14 248 8 691 VL 0,5 141,80885 1,42 12,33 45 Senaki 52 112 31 788 VL 0,5 44,26518 0,44 14,07 46 Chkhorotsku 30 124 18 376 VL 0,5 6,84251 0,07 1,26 47 Tsalenjikha 40 133 24 481 VL 0,5 23,36586 0,23 5,72 48 Khobi 41 240 25 156 VL 0,5 25,80636 0,26 6,49 49 Adigeni 20 752 12 659 VL 0,5 61,54544 0,62 7,79 50 Aspindza 13 010 7 936 VL 0,5 166,178 1,66 13,19 51 Akhalqalaqi 60 975 37 195 VL 0,5 54,44271 0,54 20,25 52 Akhaltsikhe 46 134 28 142 VL 0,5 199,06086 1,99 56,02 53 Borjomi 32 422 19 777 VL 0,5 73,02162 0,73 14,44 54 Ninotsminda 34 305 20 926 VL 0,5 7,92003 0,08 1,66 55 Rustavi 116 384 70 994 M 1,5 522,15781 5,22 370,7 58

56 Bolnisi 74 301 45 324 VL 0,5 96,58615 0,97 43,78 57 Gardabani 114 348 69 752 VL 0,5 82,37407 0,82 57,46 58 Dmanisi 28 934 17 650 VL 0,5 11,8242 0,12 2,09 59 Tetri Tskaro 25 354 15 466 VL 0,5 29,63072 0,30 4,58 60 Marneuli 118 221 72 115 VL 0,5 83,41945 0,83 60,16 61 Tsalka 20 888 12 742 VL 0,5 60,40063 0,60 7,7 62 Gori 148 686 90 698 H 3,5 1712,56543 1,71 1418 63 Kaspi 52 217 31 852 VL 0,5 44,40598 0,44 14,14 64 Kareli 50 422 30 757 VL 0,5 41,91836 0,42 12,89 65 Khashuri 62 714 38 256 VL 0,5 56,10204 0,56 21,46

DISCUSSION This is the first time the multiplier/benchmark method has been applied to estimate an IDU population in Georgia.

While every effort has been made to produce as accurate an indication of the prevalence of drug use as possible, these estimates are based on sparse data of poor quality. It is clear that more robust estimates of the size of the drug using population are required. This can only be done through indirect prevalence estimation, using a range of methods and data sources. It is also clear, however, that the secondary data necessary for such estimation, and the capacity to undertake it, is lacking in Georgia.

There was wide variation in the estimates derived from the different multipliers. There may have been some inflation of the treatment prevalence estimates because treatment data would have included a small number of duplicate episodes where a person has been transferred between services. The prevalence estimates for IDUs that were derived using the benchmark data from low threshold services were higher than those derived using other multipliers. Three of the estimates derived from the various multipliers seem more realistic: one derived from the police data, one derived from data on HIV tests, and one derived from the proportion that had been in treatment (unfortunately this indicator is available only in 2 cities of Georgia). The estimate derived from the low threshold agencies may be worth focusing on, as the collection of this data may be the most robust of all the routine sources of data, since some experts suggest that estimates based on these data sources must be an overestimate.

Each indicator selected to calculate the IDU estimates has biases; each indicator that we considered in this study is based on a different way of “encountering” an IDU. HIV counseling and testing and drug abuse treatment are usually based on voluntary interaction with health agencies. Data on treatment demand and HIV testing and counseling events depend on the desires of potential clients and on the availability of capacity at the service agency, they can happen multiple times a year for some persons and much less often for others. Drug abuse treatment and HIV counseling and testing services may be funded more or less adequately, and this can change over time. HIV testing and counseling encounters also depend on the physical locations of sites where these services are provided.

Regardless of its size it is apparent that there is a population of IDUs in Georgia that is currently underserved by the health sector. Implementing a broad range of health services for drug users and strengthening the data collection capabilities of the providers of these services would help to generate the data necessary for indirect estimation.

On the other hand, the multiplier method used in this study has its advantages. Firstly, the result suggests that combining this method with the HIV/AIDS behavioural surveillance to produce population size estimations is feasible and cost effective – in this way the necessary parameters for the estimation can be simply obtained. Secondly, combining this method with the BSS, estimates can be obtained regularly (under the framework of the National Surveillance System) and trends in the size of IDU populations with time can be observed. Furthermore, this method can be generalized to the whole BSSs, and thus estimates can be obtained for broader geographical areas.

The methodology used for recruiting IDUs in the study - RDS offers certain unique features (Heckathorn 1997). It reduces the biases associated with non-random recruitment, allows greater penetration into diverse groups of IDUs, and allows respondents to recruit only a limited number of respondents irrespective of their network size (Magnani et al 2005). One advantage of the RDS method is that the sampling frame is built up during the recruitment process and this helps to avoid 60 incomplete sampling frames. In this manner it provides unbiased population estimates. Additionally, at the stage of analysis, RDS takes into account, the different network-sizes and to what extent each respondent has recruited others like him/her. Another theoretical advantage of RDS is that it is based on a dual incentive system, financial rewards in combination with peer pressure, which can be expected to reduce non-response bias. All these factors make RDS a superior method for recruitment as compared to conventional methods like snow-ball sampling.

The study was conducted using minimum of resources. NGO already working on the ground implemented the study. Additionally, the staff members experienced in BSS had been involved in data collection. Thus, no new structures were required to be established. Moreover, the recruitment process, using RDS, was surprisingly fast - the data collection was completed within just 10-14 days in each site.

Some key issues must be kept in mind in using multiplier methods successfully for IDU population size estimation. Firstly, a clear and consistent definition of IDUs in different surveys should be used. Even when referring to the broadest possible target group, the „drug users“, any definition should include: a time period, an age group, frequency of use, and a definition of substances. Secondly, the catchment area for the selected data sources should be ideally the same as that covered in the survey from which multipliers are derived.

Possible limitations to the study could have affected the results. The small numbers of women participating in the surveillance may indicate a strong desire to remain hidden, their limited numbers, or a reflection of poor recruiting. Because few women have been arrested or attended treatment facilities, there are only some data regarding injection drug use amongst women in Georgia.

Reporting bias: as in any interview-based surveys, it is possible that respondents may not have accurately answered some of the sensitive questions, or may have had difficulties in recalling information.

The applicability of the Multiple Indicator Method for the extrapolation from local to national prevalence estimates as proposed by the EMCDDA was of limited use in the Georgian context because of a lack of drug-related indicators throughout the country. Among anchor points the prevalence estimation was derived based on the limited number of indicators – all 5 indicators were available only in 2 cities. Single drug related indicator such as number of IDUs registered by Police for drug consumption could be obtained even in 13 cities. Among the demographic indicators only population density was available.

The prevalence estimates that are used as anchor points in a multiple indicator analysis will have an impact on the prevalence figures derived for other areas. These anchor points must be available for at least two of the areas (preferably far more than two areas) and must be valid and reliable as they determine the parameters of the regression model.

Generally, the analysis showed that the total prevalence is highly dependent on the choice of anchor points, as these anchor points give the actual span of prevalence between which the regions are spread. Therefore, anchor points should be from both sides of the continuum, i.e. the estimates should cover at least one area with an assumed high prevalence rate and one region at the lower end of prevalence rates, in order to improve the quality of the regression model. Usually, however, there are very few anchor points from which to extrapolate, which results in highly unstable multiple regression predictions, and so an amended regression procedure is required.

Obviously, it is easier to get high prevalent anchor points than low prevalent anchor points since scientific projects are more often conducted in regions where the drug problem has become apparent. The same situation was appeared in Georgia – out of five anchor points none of them represented the low prevalence area.

Reliability and validity of estimates for the anchor points are of critical importance. On the other hand, the unobserved prevalence is related to the observed indicators, and that the relationship between the indicators and the anchor points is similar for other geographical areas. However, other factors also have a bearing on the indicators and may invalidate that assumption and the derived results, in particular, the number of drug users in treatment may be restricted by the capacity of treatment services, or affected by the level of underreporting that can vary across the country; otherwise the level of policing and attention given to drugs offences may vary across the country and etc. However, research has shown, that the anchor points have a much greater impact on the national prevalence estimate than the choice of indicators.26

The total prevalence estimates derived by the different calculations reached from around 39 000 IDUs to 41 000 which is a difference of no more than 2 000. Comparison made of the extrapolated - predicted values of estimated drug prevalence derived from the regression model for the anchor points themselves with the actual data provided for those points revealed that there is a considerable difference between the anchor point observed prevalence and the prevalence predictions of the multiple indicator model. This might be caused by the lack of appropriate original data in target points.

26 Prevalence of problem drug use at the national level, EMCDDA, 2002

CONCLUSION AND RECOMMENDATIONS

Understanding something about the dynamics of the drug problem makes it possible not only to assess the likely impact of the problem, but also to alert policy makers to a worsening situation, or alternatively to provide evidence that prevention and other initiatives may be working. Although the need for information on the scale of the drug problem is clear, the data are, in practice, extremely hard to generate. Of all the methods of indirect estimation the multiplier-benchmark approach is probably the easiest to implement and probably the one with the longest history of use in the field of drug epidemiology.

Given that the concordance of different methods probably gives the best indicator of a satisfactory estimate being derived, prevalence estimates derived from a range of methods should be obtained and the different estimates compared and contrasted to help in selecting the “best estimate”. We should use both capture-recapture and multiplier-benchmark methods if possible; Because of problems in obtaining data the capture-recapture method cannot be used in Georgia. When using multiplier/benchmark method, multiple multipliers generated from more than one source should be applied.

In this report we have provided the first ever estimates of the prevalence of injection drug use within Georgia. We have shown that injection drug use is occurring in all selected cities of Georgia and that, on average, from 1, 30% to 7, 97% of the population aged between 18 and 64 has used these drugs within the year 2007. Since this is the first time that an estimate of this kind has been produced there is no previous figure with which to make comparisons. The current study has demonstrated that it is feasible to apply the multiplier-benchmark method to the task of estimating the size of IDU population in Georgia.

The multiple indicator method to derive national prevalence estimates is cost-effective, as it does not require new data collection, unless separate studies are needed to estimate new anchor points. Evidently, increasing the number of anchor points makes the regression more stable. Local estimation methods should be used and further developed to produce regional anchor points for the multivariative indicator method.

Only prevalence estimates of cities, but not of the surrounding regions are available in Georgia. As problem drug use is concentrated heavily in these cities they may be used as anchor points. If only cities are employed as anchor points the requirement is that also low prevalent regions should be available. One problem encountered during the study was obtaining data across the country to use a variety of drug indicators available in the anchor and target areas in order to improve the predictability of drug abuse prevalence nationwide. The current study suggests that a more differentiated response to the problem of drug abuse may be possible, although more work is required to provide more detailed breakdowns in terms of drug-related and demographic indicators.

REFERENCES Developing an HIV Behavioral Surveillance System for Injecting Drug Users: The National HIV Behavioral Surveillance System. Amy Lansky et al, Public Health Reports/2007 Supplement 1/Volume 122 Behavioral Surveillance Surveys (BSS): Guidelines for Repeated Behavioral Surveys in Populations at Risk of HIV. Family Health International, 2000 Integrated HIV Behavioral and Serologic Surveillance System, Manual of Procedures. FHI, 2006 Sampling Hard-to-Reach Populations. Population Services International, 2007 Sampling and Estimation in Hidden Populations Using Respondent-Driven Sampling. Matthew J. Salganik and Douglas D. Heckathorn. Sociological Methodology, 2004 Review of sampling hard-to-reach and hidden populations for HIV surveillance. Robert Magnani, Keith Sabin, Tobi Saidel and Douglas Heckathorn. AIDS 2005, 19 (suppl 2) Street and Network Sampling in Evaluation Studies of HIV Risk-Reduction Interventions. Salaam Semaan, Jennifer Lauby and Jon Liebman. AIDS Review 2002; 4 Methodological Guidelines to Estimate the Prevalence of Problem Drug Use on the Local Level. EMCDDA, December 1999 Estimating the Prevalence of Problem Drug Use in Europe. EMCDDA Scientific Monograph Series N1 Prevalence of problem drug use at the national level. EMCDDA, 2002 Key Epidemiological Indicator: Prevalence of problem drug use. EMCDDA, 2004 Approaches to Estimating Drug Prevalence in Ireland: An Overview of Methods and Data Sources. National Advisory Committee on Drugs, 2003 Estimating Prevalence: Indirect Methods for Estimating the Size of the Drug Problem. Global Assessment Programme on Drug Abuse, Toolkit Module 2. UNODC, 2003 Estimating the Size of Populations at Risk for HIV: Issues and Methods. UNAIDS and FHI, 2003 HIV in San Francisco: Estimated Size of Populations at Risk, HIV Prevalence and HIV Incidence for 2006. Willi McFarland, CDC, 2007 Estimating the number of drug injectors in Indonesia. Elizabeth Pisani, International Journal of Drug Policy N 17, 2006 Size Estimation of Injecting Drug Use in Punjab and Haryana. Atul Ambekar and B. M. Tripathi, UNAIDS 2008 Estimating Local and National Problem Drug Use prevalence from Demographics, Filip Smit et al., Addiction Research and Theory, 2003, Vol. 11, N 6 Estimating numbers of IDUs in metropolitan areas for structural analyses of community vulnerability and for assessing relative degrees of service provision for IDUs. Samuel R. Friedman et al. Journal of Urban Health, 2004 Characteristics, High-Risk Behaviors and Knowledge of STI/HIV/AIDS, and HIV and Syphilis Prevalence among Injecting Drug Users in Tbilisi, Georgia – 2002. Save the Children: STI/HIV Prevention (SHIP) Project, 2003 The system of registration and follow-up of drug users in Georgia. David Otiashvili and Nino Balanchivadze, Addiction Research Center Alternative Georgia, 2008 Anti drug Strategy of Georgia. Tamar Sirbiladze et al, MoLHSA and Bemoni Public Union, 2006

APPENDIXES

Appendix 1. Sociodemographic characteristics of nonrespondents in research studies using nomination techniques

Table 31. Characteristics of non-respondents in IDU prevalence study

Tbilisi Gori Telavi Zugdidi Batumi Non- Sample Non- Sample Non- Sample Non- Sample Non- Sample respondents respondents respondents respondents respondents n % n % n % n % n % n % n % n % n % n %

12 3.9 307 100 14 6.8 205 100 26 12.7 205 100 26 12.7 204 100 9 4.4 206 100 Age 18-24 1 8,3 21 6.8 2 14.3 35 17.1 5 19,2 34 16.6 2 7,7 27 13.2 3 33,3 25 12.1 25-30 1 8,3 51 16.6 3 21,4 38 18.5 8 30,8 56 27.3 4 15,4 45 22.1 2 22,2 47 22.8 31-40 3 25,0 80 26.1 5 35,7 74 36.1 9 34,6 75 36.6 11 42,3 78 38.2 1 11,1 78 37.9 41-50 6 50,0 122 39.7 4 28,6 50 24.4 4 15,4 36 17.6 6 23,1 40 19.6 2 22,2 48 23.3 50+ 1 8,3 33 10.7 8 3.9 4 2.0 3 11,5 14 6.9 1 11,1 8 3.9 Gender Male 12 100 304 99 14 100 200 97.6 26 100 205 100 26 100 203 99.5 9 100 200 97 Female 3 1 5 2.4 1 0.5 6 3 Marital Status Married 7 58,3 167 54.4 8 57,1 111 54.1 8 30,8 85 41.5 12 46,2 106 52.0 5 55,6 99 48.1 Divorced 3 25,0 62 20.2 1 7,1 20 9.8 1 3,8 21 10.2 4 15,4 8 3.9 27 13.1 Has never 5 35,7 17 65,4 10 38,5 4 44,4 78 25.4 74 36.1 99 48.3 90 44.1 80 38.8 been married 2 16,7 Education Level Secondary 3 25,0 82 26.7 10 71.4 132 67.3 22 84,6 139 67.8 21 80,8 118 57.8 8 88,9 142 68.9 Incomplete 1 8,3 1 3,8 1 3,8 16 5.2 7 3.4 5 2.4 12 5.9 15 7.3 High High 8 66,7 209 68.1 4 28.6 60 29.3 3 11,5 61 29.8 4 15,4 74 36.3 1 11,1 49 23.8

Appendix 2. Key characteristics of IDU sample in five locations across the country

Table 32. Key Characteristics of IDU sample in Tbilisi

RDS population Adjusted Characteristics estimates, n/N n/N % (95% Cl) Socio-demographic characteristics Age 18-24 7.6 (4-12) 21/307 20/300 25-30 16.8 (12.3 -21.4) 51/307 50/300 31-40 26.5 (21.3 -31.7) 80/307 78/300 41-50 38.8 (32 – 46) 122/307 120/300 50+ 10.3 (7 -14) 33/307 32/300 Mean 38.4 Median 40 Gender Male 99.3 (98.1 – 99.7) 304/307 298/300 Female 0.7 (0/3 – 1.7) 3/307 2/300 Educational level Secondary 27.1 (22 – 32.7) 82/307 81/300 Incomplete high 5.3 (3 -8 ) 16/307 16/300 High 67.6 (62 – 73) 209/307 203/300 Marital status Married 53.6 (47 – 60.3) 167/307 164/300 Divorced/Separated 21 (16.3 – 25.7) 62/307 61/300 Never been married 25.4 (20.3 – 31) 78/307 75/300 Drug use history Age when first used any drug <15 24.9 (20 – 30.3) 76/307 74/300 15-19 62.9 ( 57 – 68.7) 192/307 189/300 20-24 10.2 (7 – 13.7) 33/307 31/300 25+ 1.9 (0.7 – 3.7) 6/307 6/300 Mean (minimum – maximum) 16.33 ( 9 – 35) Median 16 Age when first injected drugs <15 4.6 (2.3 – 7) 14/307 14/300 15-19 52.9 (46.7 – 59) 163/307 159/300 20-24 26.7 (21 – 32.4) 83/307 80/300 25+ 15.8 ( 11.3 – 20.7) 4/307 47/300 Mean (minimum – maximum) 19.81 (13 – 35) Median 19 Duration of injecting drug use in years Mean (minimum – maximum) 11.93 (0.4 – 37) Median 10 Frequency of injecting drug use in the last week Once a week 6.6 (3.7 – 9.7) 20/307 19/300 Several times a week 43.7 (38 – 49.4) 130/307 129/300 67

Once a day 18.5 (14.3 -22.7) 57/307 57/300 Several times a day 13.8 (9.7 – 17.7) 42/307 42/300 Have not taken 17.0 (13 – 22) 57/307 52/300 Member of regular injecting group Yes 81.3 (75.3 – 86.7) 253/307 246/300 No 18.7 (13.3 – 24.7) 54/307 54/300 Mean number of injecting group members 4.67 (2 – 15) Consumed drugs last week 51.6 (47 – 56.3) 161/307 155/300 CNS depressants 47.5 (39.3 – 55.4) 76/161 73/155 CNS stimulants 0 Narcotic drugs 41.5 (33.3 – 49.7) 67/161 65/155 Hallucinogens 45.1 (36.8 – 53.5) 73/161 69/155 Other psychoactive substances 3.8 (1.3 – 6.9) 7/161 6/155 Mean # of drugs used 1.66 (1-5) Injected drugs last week 82.9 (78.3 – 87.3) 250/307 248/300 CNS depressants 2.6 (0.8 – 5.2) 7/250 7/248 CNS stimulants 15.7 (10.8 – 20.9) 39/250 39/248 Narcotic drugs 61.7 (55.2 – 68.1) 154/250 152/248 Hallucinogens 0 Other psychoactive substances 7.9 (4.7 – 11.5) 20/250 20/248 Combination 41.9 (35.4 – 48.5) 104/250 104/248 Mean # of drugs used 1.46 (1-4) Exposure to interventions Drug treatment Currently taking medical treatment. 3.2 (1.3 – 5.3) 12/307 10/300 Used to take medical treatment, but quit 39.4 (34 – 46.3) 124/307 122/300 Never have been treated 57.3 (50.5 – 64.2) 171/307 168/300 Other services 5.1 (3– 7.7) 14/300 Had voluntary HIV test and received results 16/307

IDUs who where given condoms in the last 12 8.9 (6 – 12) 27/307 27/300 months IDUs who where given brochures/pamphlets/ 33.9 (29 – 39) 107/307 101/300 booklets on HIV/AIDS in the last 12 months IDUs who where given qualified information 19.6 (15.3 – 24) 63/307 58/300 on HIV/AIDS in the last 12 months I2.3 IDUs who have heard/seen/read information about syringe exchange program in the last 14.9 (11 – 18.7) 46/307 45/300 12 months IDUs who where given sterile syringes in the 4.0 (1.7 – 6.6) 12/307 12/300 last 12 months IDUs who where given information about substitution therapy program in the last 12 93.4 (89.7 – 96.7) 285/307 279/300 months IDUs who used substitution therapy program 1.9 (0.3 – 3.7) 6/307 6/300 in the last 12 months Biomarker 2.5 (0.3 – 5.4) 7/306 7/299 HIV

Syphilis 6.3 (3.7 – 9.3) 19/306 19/299

Table 33. Key Characteristics of IDU sample in Gori

RDS population Adjusted Characteristics estimates, n/N n/N % (95% Cl) Socio-demographic characteristics Age 18-24 23.1 (14.1 -33) 35/205 35/200 25-30 20.0 (14.5 – 25.9) 38/205 38/200 31-40 33.6 (26.6 – 41) 74 /205 73/200 41-50 18.4 (11.9 – 25.6) 50/205 46/200 50+ 4.9 (2 – 8.4) 8 /205 8/200 Mean 34.57 Median 34 Gender Male 97.7 (95 – 99.5) 200/205 195/200 Female 2.3 (0.5 -5) 5/205 5/200 Educational level Primary 0.9 (0 – 2.5) 2 /205 2/200 Secondary 67.1 (59.5 – 74.5) 136 /205 133/200 Incomplete high 3 (1 -5.5) 7 /205 6/200 High 28.9 (22.1 -36) 60 /205 59/200 Marital status Married 54.1 (47.5 – 61) 109 /205 107/200 Divorced/Separated 9.2 (5.5 – 13.5) 20 /205 19/200 Widower 0.5 (0 – 1.5) 2 /205 1/200 Never been married 36.2 (30 – 42.5) 74 /205 73/200 Drug use history Age when first used any drug <15 11.8 (7.5 – 16.4) 24/205 24/200 15-19 67.9 (61 – 74.9) 140/205 135/200 20-24 16 (11 – 21.5) 31/205 31/200 25+ 4.4 (2-7.1) 10/205 10/200 Mean (minimum – maximum) 17,57 (12-39) Median 17 Age when first injected drugs <15 0.07 (0.05 – 3.5) 2/205 2/200 15-19 39.7 (32.5 – 47.5) 84/205 80/200 20-24 35.8 (29-42) 72/205 71/200 25+ 23.8 (17.5 – 30) 47/205 47/200 Mean (minimum – maximum) 21.96 (14-53) Median 20 Duration of injecting drug use in years Mean (minimum – maximum) 7.42 (0.2 – 39) Median 5 Frequency of injecting drug use in the last week Once a week 10.6 ( 6 – 15.4) 22/205 22/200 Several times a week 33.6 (28.569 – 40.4) 67/205 65/200

Once a day 6.1 (3-9) 12/205 12/200 Several times a day 5.1 (2.5 -8) 11/205 10/200 Have not taken 43.4 (36 – 50.9) 91/205 89/200 Member of regular injecting group Yes 75.6 (69 – 81.9) 151/205 150/200 No 24.4 (18.1 – 31) 54/205 50/200 Mean number of injecting group members 4.62 (2 – 15) Consumed drugs last week 16.6 (11.5 -22) 33/205 33/200 CNS depressants 51.7 (31 – 72.4) 17/33 17/33 CNS stimulant 0 Narcotic drugs 4.1 (0 -10) 1/33 1/33 Hallucinogens 51.7 (31.6 – 72.2) 17/33 17/33 Other psychoactive substances 11.8 (0 – 20) 3/33 3/33 Mean # of drugs used 1.27 Injected drugs last week 55. 1 (48 – 62.5) 114/205 111/200 CNS depressants 0 CNS stimulant 45.6 (35.9 – 55.4) 51/114 51/111 Narcotic drugs 47.2 (36.4 – 57.9) 55/114 52/111 Hallucinogens 0 Other psychoactive substances 2.7 (0.9 – 6.1) 3/114 3/111 Combination 15.7 (8.1 – 24.5) 17/114 17/111 Mean # of drugs used 1.17 Exposure to interventions Drug treatment Currently taking medical treatment. 0 Used to take medical treatment, but quit 15.2 (10.4 – 20.5) 33/205 31/200 Never have been treated 84.8 (79.5 – 89.6) 172/205 169/200 Other services Had voluntary HIV test and received results 8.4 (4.9 – 12.5) 20/205 19/200 IDUs who where given condoms in the last 12 20.3 (14.5 – 26.5) 47/205 42/200 months IDUs who where given brochures/pamphlets/ 25.4 (18.5 – 32.5) 59/205 55/200 booklets on HIV/AIDS in the last 12 months IDUs who where given qualified information 20.6 (14-27.5) 48/205 45/200 on HIV/AIDS in the last 12 months I2.3 IDUs who have heard/seen/read information about syringe exchange program in the last 49.5 (42.4 – 57) 105/205 100/200 12 months IDUs who where given sterile syringes in the 18.3 (12- 25) 44/205 40/200 last 12 months IDUs who where given information about substitution therapy program in the last 12 57.4 (50.5 – 64.5) 120/205 115/200 months IDUs who used substitution therapy program 0 in the last 12 months Biomarker HIV 0 Syphilis 3.9 (1.1 – 7.3) 7/187 7/183

Table 34. Key Characteristics of IDU sample in Telavi

RDS population Adjusted Characteristics estimates, n/N n/N % (95% Cl) Socio-demographic characteristics Age 18-24 18.3 (12.5 – 24.5) 34/205 33/200 25-30 29.5 (22.3 – 37.2) 55/205 54/200 31-40 35.8 (28.5 – 43.2) 76/205 73/200 41-50 14.9 (9 – 21.6) 36/205 36/200 50+ 1.5 (0 – 4) 4/205 4/200 Mean 33 Median 32 Gender Male 100 205/205 200/200 Female 0 Educational level Primary 1.9 (-- – --) 1/205 1/200 Secondary 64.4 (60.5 – 75) 138/205 134/200 Incomplete high 3.8 (0.5 – 4.5) 5/205 5/200 High 30 (22 – 36.5) 61/205 60/200 Marital status Married 41 (33.5 – 49.6) 84/205 82/200 Divorced/Separated 10.4 (6 – 15) 21/205 21/200 Widower 0.7 (0 – 1.5) 1/205 1/200 Never been married 47.9 (39.5 – 56) 99/205 96/200 Drug use history Age when first used any drug <15 12.3 (7.5 – 17.5) 26/205 25/200 15-19 67.9 (61.5 – 74.1) 139/205 135/200 20-24 14.4 (1 – 19) 29/205 29/200 25+ 5.4 (2.5 – 8.5) 11/205 11/200 Mean (minimum – maximum) 17.5 (13-40) Median 17 Age when first injected drugs <15 0.5 (0 – 1.5) 1/205 1/200 15-19 35 (29.5 – 40.6) 74/205 71/200 20-24 37.8 (32.1 – 43.5) 77/205 75/200 25+ 26.6 (20.9 – 32.5) 53/205 53/200 Mean (minimum – maximum) 21.9 (14-42) Median 20 Duration of injecting drug use in years Mean (minimum – maximum) 8.2 (0.5-30) Median 6 Frequency of injecting drug use in the last week Once a week 9.8 (6 – 13.5) 19/205 19/200 Several times a week 8.7 (4.5 – 13)71 17/205 17/200

Once a day 0 Several times a day 1.4 (0 – 2.5) 2/205 2/200 Have not taken 80.1 (75.4 – 87) 167/205 162/200 Member of regular injecting group Yes 67 (60.5 – 73.5) 136/205 134/200 No 33 (26.5 – 39.5) 69/205 66/200 Mean number of injecting group members 4.5 (2-10) Consumed drugs last week 11.5 (7 – 16.5) 24/205 23/200 CNS depressants 13 (0 – 27.5) 3/24 3/23 CNS stimulant 0 Narcotic drugs 6.2 (0 – 14.8) 1/24 1/23 Hallucinogens 87 (72 – 100) 21/24 20/23 Other psychoactive substances 0 Mean # of drugs used 1.1 (1-2) Injected drugs last week 19 (13 – 25) 38/205 38/200 CNS depressants 0 CNS stimulant 36.4 (17.9 – 53.1) 14/38 14/38 Narcotic drugs 58.4 (42.5 – 75.9) 22/38 22/38 Hallucinogens 0 Other psychoactive substances 5.3 (0 – 13.9) 2/38 2/38 Combination 10.8 (2.5 – 20.8) 4/38 4/38 Mean # of drugs used 1.2 (1-3) Exposure to interventions Drug treatment Currently taking medical treatment. 0 Used to take medical treatment, but quit 29.5 (21.7 – 37.8) 66/205 64/200 Never have been treated 70.5 (62.2 – 78.4) 139/205 136/200 Other services Had voluntary HIV test and received results 2.9 (0.5 – 6) 7/205 6/200 IDUs who where given condoms in the last 12 6 (3 – 9.5) 14/205 12/200 months IDUs who where given brochures/pamphlets/ 16.4 (11 – 22) 36/205 33/200 booklets on HIV/AIDS in the last 12 months IDUs who where given qualified information 6.6 (3 – 11) 18/205 15/200 on HIV/AIDS in the last 12 months I2.3 IDUs who have heard/seen/read information about syringe exchange program in the last 25.6 (19 – 32.4) 58/205 53/200 12 months IDUs who where given sterile syringes in the 3.3 (1 – 6.5) 9/205 7/200 last 12 months IDUs who where given information about substitution therapy program in the last 12 61 (53.5 – 68) 128/205 123/200 months IDUs who used substitution therapy program 0.7 (0 – 1.5) 1/205 1/200 in the last 12 months Biomarker 1.5 (0 – 3.5) 3/200 HIV 3/205

Syphilis 5.5 (2.5 – 8.5) 11/205 11/200

Table 35. Key Characteristics of IDU sample in Zugdidi

RDS population Adjusted Characteristics estimates, n/N n/N % (95% Cl) Socio-demographic characteristics Age 18-24 15.4 (9.4-22.1) 27/204 27/200 25-30 22.6 (16.9-29) 45/204 44/200 31-40 37.9 (31-45) 78/204 78/200 41-50 18 (12.4-24) 40/204 40/200 50+ 6.1 (3-9.6) 14/204 11/200 Mean 34.8 Median 35 Gender Male 99.5 (98.5-100) 203/204 199/200 Female 0.5 (0-1.5) 1/204 1/200 Educational level Secondary 57.4 (49.6-64.5) 118/204 115/200 Incomplete high 6 (3-9.5) 12/204 12/200 High 36.6 (30-43.5) 74/204 73/200 Marital status Married 49.5 (43-57.5) 105/204 102/200 Divorced/Separated 4.6 (1.5-8.3) 8/204 8/200 Widower 0.8 (0-1.5) 1/204 1/200 Never been married 45.2 (38.5-51.5) 90/204 89/200 Drug use history Age when first used any drug <15 14.3 (9.9-19) 29/204 29/200 15-19 70.8 (65-76.5) 142/204 139/200 20-24 11 (6.5-15.9) 25/204 24/200 25+ 3.9 (1.5-6.5) 8/204 8/200 Mean (minimum – maximum) 16.8 (11-29) Median 16 Age when first injected drugs <15 2.3 (0.5-4.5) 5/204 5/200 15-19 56.2 (48.5-64) 110/204 108/200 20-24 24.3 (18-30.6) 51/204 49/200 25+ 17.2 (12-22.5) 38/204 38/200 Mean (minimum – maximum) 20 (14-40) Median 18 Duration of injecting drug use in years Mean (minimum – maximum) 9.8 (0.5-30) Median 8 Frequency of injecting drug use in the last week Once a week 7.8 (4-11) 17/204 16/200 Several times a week 10.6 (6-15.1) 19/204 19/200 Once a day 2.1 (0-3.5) 73 2/204 2/200

Several times a day 2.3 (0.5-4) 4/204 4/200 Have not taken 77.3 (73-84.4) 162/204 159/200 Member of regular injecting group Yes 64 (57.5-70.5) 132/204 128/200 No 36 (29.5-42.5) 72/204 72/200 Mean number of injecting group members 4.6 (2-15) Consumed drugs last week 8 (4.5-12) 16/204 16/200 CNS depressants 8.7 (0-21.1) 1/16 1/16 CNS stimulant 0 Narcotic drugs 0 Hallucinogens 75 (50-94.4) 12/16 12/16 Other psychoactive substances 18.6 (0-41.7) 3/16 3/16 Mean # of drugs used 1 (1-1) Injected drugs last week 21.2 (16-26.5) 44/204 43/200 CNS depressants 0 CNS stimulant 23.6 (11.8-35.9) 10/44 10/43 Narcotic drugs 62 (46.2-78) 28/44 27/43 Hallucinogens 0 Other psychoactive substances 15.9 (2.6-22.2) 5/44 5/43 Combination 18.1 (6.1-31.9) 8/44 8/43 Mean # of drugs used 1.2 (1-2) Exposure to interventions Drug treatment Currently taking medical treatment. 0.7 (0-1.5) 1/204 1/200 Used to take medical treatment, but quit 37.5 (31-44.5) 76/204 75/200 Never have been treated 61.8 (55.5-68.6) 127/204 124/200 Other services Had voluntary HIV test and received results 5.2 (2.5-8) 10/204 10/200 IDUs who where given condoms in the last 12 33/204 32/200 months 16.1 (11.5-21) IDUs who where given brochures/pamphlets/ 33.5 (27.3-40) 69/204 66/200 booklets on HIV/AIDS in the last 12 months IDUs who where given qualified information 11.8 (7-17) 24/204 24/200 on HIV/AIDS in the last 12 months I2.3 IDUs who have heard/seen/read information about syringe exchange program in the last 24.8 (18-32) 50/204 49/200 12 months IDUs who where given sterile syringes in the 1.1 (0-2.6) 2/204 2/200 last 12 months IDUs who where given information about substitution therapy program in the last 12 63.2 (55.6-70.5) 130/204 127/200 months IDUs who used substitution therapy program 63.2 (55.5-70.5) 130/204 130/200 in the last 12 months Biomarker HIV 2.2 (0-3.5) 3/204 3/200 Syphilis 6.9 (3.5-11) 14/204 14/200

Table 36. Key Characteristics of IDU sample in Batumi

RDS population Adjusted Characteristics estimates, n/N n/N % (95% Cl) Socio-demographic characteristics Age 18-24 12.5 (7-18.8) 25/206 24/200 25-30 24.6 (17-33.2) 47/206 45/200 31-40 37.5 (30-45.2) 78/206 76/200 41-50 22.1 (15.9-28.5) 48/206 47/200 50+ 3.3 (1-6) 8/206 8/200 Mean 35.07 Median 35 Gender Male 98.1 (95.5-100) 200/206 195/200 Female 1.9 (0-4.5) 6/206 5/200 Educational level Secondary 70.2 (63-77) 142/206 140/200 Incomplete high 7 (3.5-10.9) 15/206 14/200 High 22.8 (16.5-29.5) 49/206 46/200 Marital status Married 47.2 (39.5-54.7) 95/206 94/200 Divorced/Separated 12.4 (7.5-17.6) 27/206 25/200 Widower 1.9 (0.5-4) 4/206 4/200 Never been married 38.5 (31-46) 80/206 77/200 Drug use history Age when first used any drug <15 28.4 (21.6-35.5) 59/206 56/200 15-19 59 (52-66) 120/206 118/200 20-24 10.3 (6-14.9) 22/206 21/200 25+ 2.3 (0.5-4.5) 5/206 5/200 Mean (minimum – maximum) 16.2 (9-30) Median 16 Age when first injected drugs <15 2.7 (1-5) 6/206 6/200 15-19 53.9 (46-62) 110/206 106/200 20-24 27.5 (21-34.5) 57/206 56/200 25+ 15.9 (10.5-22) 33/206 32/200 Mean (minimum – maximum) 19.8 (13-34) Median 19 Duration of injecting drug use in years Mean (minimum – maximum) 9.9 (0.5-40) Median 7 Frequency of injecting drug use in the last week Once a week 11.3 (7-15.5) 23/206 23/200 Several times a week 42.5 (33.8-51.5) 82/206 81/200 Once a day 3.2 (1-5.6) 75 7/206 6/200 Several times a day 4.2 (1.5-7.5) 10/206 10/200 Have not taken 38.8 (30.7-47) 84/206 80/200 Member of regular injecting group Yes 60.1 (53-67) 124/206 120/200 No 39.9 (33-47) 82/206 80/200 Mean number of injecting group members 3.8 (2-10) Consumed drugs last week 23.9 (18-30.1) 49/206 48/200 CNS depressants 34 (20-48.9) 16/49 16/48 CNS stimulant 6.1 (0-11.5) 2/49 2/48 Narcotic drugs 29.5 (16.2-43.5) 15/49 14/48 Hallucinogens 47.8 (33.3-62.3) 23/49 23/48 Other psychoactive substances 4.1 (0-10.2) 2/49 2/49 Mean # of drugs used 1.3 (1-3) Injected drugs last week 61.1 (52.9-69) 122/206 119/200 CNS depressants 0.9 (0-2.9) 1/122 1/119 CNS stimulant 13 (6.5-20.2) 16/122 15/119 Narcotic drugs 91.4 (85.7-96.4) 112/122 110/119 Hallucinogens 0 Other psychoactive substances 0 Combination 4.2 (0.9-7.8) 5/122 5/119 Mean # of drugs used 1.1 (1-4) Exposure to interventions Drug treatment Currently taking medical treatment. 4.5 (2-7.5) 9/206 9/200 Used to take medical treatment, but quit 43.9 (36.5-51.5) 90/206 88/200 Never have been treated 51.6 (44-59.1) 107/206 103/200 Other services Had voluntary HIV test and received results 4.2 (1.5-7.5) 12/206 10/200 IDUs who where given condoms in the last 12 25.6 (19.5-32) 57/206 52/200 months IDUs who where given brochures/pamphlets/ 29.2 (22.5-36) 67/206 61/200 booklets on HIV/AIDS in the last 12 months IDUs who where given qualified information 21.5 (16-27.5) 48/206 43/200 on HIV/AIDS in the last 12 months I2.3 IDUs who have heard/seen/read information about syringe exchange program in the last 42.7 (35-50.5) 91/206 87/200 12 months IDUs who where given sterile syringes in the 8.1 (4-12.5) 21/206 18/200 last 12 months IDUs who where given information about substitution therapy program in the last 12 81.1 (75.5-86.5) 168/206 162/200 months IDUs who used substitution therapy program 1.4 (0-2.5) 2/206 2/200 in the last 12 months Biomarker HIV 4.5 (1.5-8) 9/206 9/200 Syphilis 7.6 (4-12) 15/206 15/200

Appendix 3. Nomination method/questionnaire

Questionnaire Identification Number:

Coupon Number:

1. What is the number of your close friends with whom you have been using drugs in 2007 (or whom you know for sure they are or were using drugs, including those who passed away and those who ceased to use drugs meanwhile)?

2. Are you sure? Could you please think about this number for me for a while? Sounds to me (too high or low /too quick/ too round). Maybe you could name them by their first names (even unreal, imaginary) to obtain more specific number? Names: I. ______II. _____ III. _____ IV. _____ V. ______Final number:

3. Was (name) ____ tested by police for presence of illegal drugs in 2007? 1. Yes 2. No 88. Don’t know 99. No response

4. Was (name) ____ tested for HIV in 2007? 1. Yes 2. No 88. Don’t know 99. No response

5. Was (name) ____ in abstinence-oriented treatment in 2007? 1. Yes (Go to Q. 8) 2. No 88. Don’t know Continue 99. No response

6. Was (name) ____ considering entering the abstinence oriented treatment in 2007, but did not do so? 1. Yes (Continue) 2. No (Go to Q.8) 88. Don’t know (Continue) 99. No response (Continue)

7. Why s/he did not? 1. Changed his mind 2. Because of high cost 77

3. Entered the substitution treatment 4. Any other reason 88. Don’t know 99. No response

8. Was (name) ____ in substitution treatment in 2007? 1. Yes (Go to Q. 10) 2. No 88. Don’t know Continue 99. No response

9. Was (name) ____ in substitution treatment waiting lists in 2007? 1. Yes 2. No 88. Don’t know 99. No response

10. Was (name) ____ in the needle exchange (when used needles are changed by new ones) and other low-threshold programs (e.g. voluntary counseling and testing on Hepatitis B, C and HIV/AIDS by physicians and psychologists) in 2007? 1. Yes 2. No 88. Don’t know 99. No response

11. Was (name) ____ deceased due to a fatal drug overdose in 2007? 1. Yes 2. No 88. Don’t know 99. No response

Questions 3-11 will be asked for every nominated drug user.

Thank you indeed!